The merchandise trade statistics measure goods traded between the United
States and other countries. They are the official source of information
about U.S. imports, exports and balance of merchandise trade. As a leading
economic indicator and a major component of the Gross Domestic Product
(GDP), the statistics provide critical information to a wide and varied
group of users in the public and private sectors. The Census Bureau strives,
with available resources, to provide accurate, high quality statistics
to meet these needs. However, to use the information wisely and appropriately,
users need to understand the nature and limitations of the merchandise
trade statistics program. This paper discusses issues affecting the quality
of the statistics and some known limitations.
Background
In 1989 the Census Bureau invited the National Academy of Sciences
(NAS) to study the merchandise trade statistics program. In its resulting
report, Behind the Numbers, released in January 1992, NAS noted that limitations
in the statistics are not well communicated to the users. This profile
addresses this concern by presenting known limitations in the trade statistics
and by evaluating known sources of error. It also includes discussions
of recent and proposed improvements in the quality of the statistics.
The Census Bureau intends to issue annual reports providing information
on any changes in the program that affect the quality of the statistics.
Comments or questions about the information in this paper or suggestions
for future reports are welcome. Address them to Kathy Puzzilla, Chief of
Methods Research and Quality Assurance Branch, Foreign Trade Division (301)
763-7760 or Bruce Walter, Assistant Chief, Foreign Trade Division (301)
763-7020.
Source of the Statistics
The Census Bureau compiles import and export statistics from information
collected by the Customs Service at ports of entry and exit. In 1992, the
Census Bureau collected trade statistics on more than 16,000,000 import
transactions and more than 13,000,000 export transactions. This information
represented shipments through more than 400 ports of entry and exit in
the United States, Puerto Rico and the Virgin Islands. Information of ninety-five
percent of the import transactions and fifty-five percent of the export
transactions was submitted electronically. The remainder was captured from
paper documents.
II. QUALITY ISSUES
Background
Users of the statistics often assume that, since the Customs Service
collects the information on import and export transactions, the published
statistics represent an exact accounting of merchandise trade. In practice
the collection of import and export statistical information is only one
of many responsibilities of the Customs Service. However, since import
information also provides the basis for the administration of tariffs,
quotas and other enforcement activities, import statistics generally represent
a better measure of trade than export statistics.
The Customs Service limits export enforcement activities to shipments exported
under export licenses. Licensed shipments represent less than five percent
of total export shipments. For the most shipments, Customs merely collects
and forwards Shipper's Export Declarations (SEDs), the export statistical
document, to the Census Bureau for processing. Few Customs ports screen
for missing SEDs or check the completeness or accuracy of information reported
on SEDs.
Sources of Error
The major sources of error in the merchandise trade statistics include
missing documentation, timeliness, and incomplete or inaccurate reporting.
A. Missing Documents
Federal regulations require importers, exporters or their agents to
submit import and export information for all merchandise shipments above
established exemption levels. Exemption levels are $1250 ($250 for quota
items) for imports and $2500 for exports. Through comparisons of trade
data with major trading partners, the U.S./Canada Data Exchange and audits
of trade documentation, the Census Bureau has identified missing information
for both import and export shipments. Missing documentation, for the reasons
previously noted, is more common with exports. The Census Bureau takes
remedial action and corrects the published statistics whenever possible.
1. Reconciliations (Trade Data Comparisons)
In the early 1970s, The United States and Canada began annual studies to
reconcile differences in merchandise trade statistics compiled by the two
countries. After adjusting for differences in trade definitions, valuation,
timing and errors, the studies uncovered a chronic understatement of each
countries's exports. The 1970 reconciliation estimated undocumented U.S.
exports to Canada at $500 million (about six percent of total estimated
exports to Canada). By 1986,
undocumented U.S. exports to Canada grew to an estimated $10.2 billion
(more than eighteen percent of total estimated exports to Canada). (See
Table A.)
In the spring of 1987 the statistical and Customs agencies from both countries
met to find a way to deal with undocumented export shipments. The agencies
agreed that the open border between the two countries and the lack of enforcement
of filing requirements for truck and rail shipments created the problem.
They knew that to introduce adequate enforcement practices would require
a larger Customs work force and impede the flow of traffic at the border.
Anticipating the adoption of the international Harmonized System of Commodity
Classification (HS) by both countries, the agencies recommended the exchange
of import information instead of collecting export documents. In July of
1987, they signed a Memorandum of Understanding that allowed the exchange
of detailed import records. In January 1990, after two years of analysis
and major program changes, the Census Bureau substituted Canadian imports
for U.S. exports. Simultaneously Canada replaced its exports with U.S.
import statistics. Both countries dropped requirements for filing export
documents.
Several immediate improvements to the U.S. export statistics resulted from
the introduction of the data exchange. Most important, it almost eliminated
undocumented shipments to Canada. Also, since the Canadian data are pre-edited,
it improved the accuracy of the data reported to Census and ensured its
inclusion in the correct statistical month.
The success of the U.S./Canadian reconciliations prompted the Census Bureau
to seek reconciliation agreements with other major trading partners. It
now conducts reconciliation studies with Japan, Korea, Mexico, Australia
and the European Economic Community (EC). Although the Bureau of Census
does not expect other data exchanges to result from these reconciliations,
the studies provide valuable information about the quality of the U.S.
merchandise trade statistics. Reconciliation studies completed with Japan,
Korea, and the EC estimate undocumented U.S. exports at between three and
seven percent (depending on the trading partner).
Specific errors in the U.S. trade data also surfaced during these studies.
For example, discrepancies between U.S. and Japanese statistics led to
the discovery that the U.S. import statistics omitted $2.4 billion in Japanese
passenger cars and trucks shipped between April 1990 and November 1992
(an adjustment appears in the December 1992 FT 900 press release issued
in March 1993).
2. Port Audits
Concern about the level of undocumented exports also led the Bureau of
Census and the Customs Service to plan a series of audits at major Customs
ports. They proposed audits at four airports, four vessel ports and two
overland ports on the Mexican border. Besides checking for undocumented
export shipments, the audits assessed the enforcement of export filing
requirements. Because of the different filing requirements for each mode
of transport (MOT), each MOT required different audit procedures. The two
agencies decided to start with airports, followed by the vessel ports and
ending with the Mexican border.
Between September 1988 and April 1989, the agencies audited air cargo manifests
and corresponding SEDs filed at Seattle/Tacoma Airport (SEATAC), Miami
International Airport (Miami), John F. Kennedy International Airport (JFK)
and Los Angeles International Airport (LAX). The audits produced estimates
of undocumented export shipments that closely paralleled the findings of
the reconciliations described earlier. The port audits estimated undocumented
exports at between two and eight percent.
The Miami audit also uncovered the presence of deliberate misclassification
and undervaluation of shipments to Central and South America. These practices
involved a small percent of export shipments to these countries. The motive
for this deliberate misreporting is the circumvention of high tariffs or
quotas in the countries of destination. The audits uncovered no evidence
of this practice at the other three airports.
The port audits proved very expensive to conduct. After the completion
of the airport audits, the agencies abandoned the audits in favor of expanding
bilateral reconciliations. These studies require fewer resources.
B. Timeliness
Timing problems result when import or export shipments are not included
in the correct transaction month. During the early 1980s scarce resources
caused the U.S. Customs Service to relax its procedures for handling statistical
documents. These changes, coupled with dramatic increases in the numbers
of import and export transactions, delayed the delivery of large numbers
of documents to the Census Bureau processing center. Often these late arrivals
were too late for inclusion in the proper month's statistics. Because these
shipments carried over into a subsequent month's statistics, the Census
Bureau coined the term "carry-over."
During the period 1984 to 1988, the Census Bureau and the Customs Service
introduced changes to collection and processing procedures to reduce the
effect of "carry-over" on the statistics. Ultimately the Census Bureau
delayed the merchandise trade release (from thirty to forty-five days after
the end of the calendar month) to allow more time for receipt and processing
of documents. In addition, it revised prior month trade totals to credit
the value of any remaining late documents to the proper statistical month.
Increased use of automated reporting by the Customs Service and the Census
Bureau also has improved the timeliness of the data and decreased "carry-over."
In 1992 "carry-over" was less than one percent of total import and export
values during initial publication of the data and less than 0.1 percent
after revision of prior month totals. (See Table B.)
C. Reporting Errors
Reporting errors are mistakes or omissions made by importers, exporters
or their agents when reporting import or export shipments. Most errors
involve missing or invalid commodity classification codes and missing or
incorrect quantities or shipping weights. They have a negligible effect
on import, export and balance of trade statistics. However, they can affect
detailed commodity and transportation statistics significantly if not corrected
or corrected inaccurately.
Recent studies show that twenty-five percent of export transactions and
eight percent of import transactions contain one or more reporting errors.
Most of these errors are easily and accurately resolved with no significant
effect on the published statistics.
Routine processing procedures include methods for correcting these errors
through clerical or electronic means. However, clerical correction is time
consuming and expensive. Defective records referred for clerical correction
can cost three times as much to process as documents that successfully
pass the edits. Also records referred for correction may not get into the
correct statistical month. Electronic corrections or imputes, which are
less expensive, may introduce some error into the statistics. Therefore,
the most accurate and cost-effective means of resolving reporting errors
is to prevent them from occurring.
Beginning in 1991 the Census Bureau introduced programs to reduce reporting
errors by educating Customs Service staff, exporters and forwarding agents
about the statistical reporting requirements.
---Customs Port Briefings
Port briefings consist of a review of Customs statistical collection
activities followed by briefings for Customs Service staff. These briefings
help increase Customs staff awareness of their role in the collection of
merchandise trade statistics. The review, conducted before the briefing,
provides background information for the briefings and frequently leads
to suggested improvements in Customs processes.
---Exporter Education Program
The goal of the Exporter Education Program (EEP) is to improve the
quality of statistics by reducing the SEDs submitted with missing, obsolete,
or invalid information. The EEP identifies exporters responsible for reporting
errors, educates them in proper reporting, tracks their improvement, and
follows-up as appropriate.
During the initial two phases of EEP, in 1991 and 1992, we concentrated
on contacting exporters responsible for specific types of reporting errors.
We mailed more than 80,000 educational packets consisting of letters, other
educational materials, and samples of defective documents filed by the
pertinent exporter. Follow-up analysis shows the program reduced reporting
errors on SEDs about twenty percent.
The current phase of EEP, which began with January 1993 statistics, takes
a different approach. Instead of targeting exporters making certain types
of errors, this phase targets 150 exporters who account for fifteen to
twenty percent of reporting errors and profiles their performance. Census
Bureau analysts contact the exporter by telephone and discuss the reporting
errors with the parties responsible for document preparation. The telephone
conversation is confirmed with a letter.
This process focuses on reducing reporting errors and by encouraging a
personal and interactive relationship between analysts and document preparation
personnel at the exporting company.
---Shipper's Export Declaration Seminars
From 1991 through 1993, the Census Bureau conducted more than thirty
free educational seminars. These seminars focus specifically on how to
prepare SEDs. More than four thousand participants from exporting companies,
forwarding agencies, carriers, and consulting firms attended. The seminars
consisted of a video presentation and question and answer sessions.
During 1993 the Census Bureau concentrated SED seminars along the Mexican
border. To meet the special needs of exporters and forwarders in the area,
it produced videos in Spanish and English on how to prepare SEDs for the
seminars. These videos may be purchased for a small fee or borrowed from
local Customs Service offices. The Census Bureau will continue programs
to educate the trade community.
D. Data Capture Errors
As explained earlier, the Census Bureau captures import and export information
from administrative documents and through various automated collection
programs. It subjects data from each source to security checks to ensure
the accurate capture of the information.
---Automated Submissions
Data received through automated collection programs; the Automated
Broker Interface (ABI), Automated Export Reporting Program (AERP) and the
U.S./Canada Data Exchange are screened upon receipt to ensure completeness
and an acceptable level of accuracy. Other security procedures guard against
file damage that may compromise the integrity of the data.
---Paper Documents
The Census Bureau uses a combination of data entry edits and keying
verification to ensure the accurate capture of information from the SEDs
and Customs Entry Forms. Automated data entry checks validate all critical
data fields and signal the keyer when the information is invalid. This
process prevents most keying errors and intercepts many reporting errors.
After data entry, the keyed information is subjected to sample verification.
Certain information common to all transactions and all information for
transactions valued over $2,000,000 for imports and $5,000,000 for exports
is 100 percent verified. Lower valued transactions are verified from a
two percent sample of documents. This provides the information needed to
assess the magnitude of error introduced through keying and the means to
intercept and correct poorly keyed batches. It also monitors the keying
performance of individual data entry clerks.
The verification plan assures an outgoing keying error rate of less than
2.10 percent for imports and 1.27 for exports. For 1990 through 1992, verification
reports showed actual outgoing error rates of about 1.25 percent for imports
and 0.60 percent for exports. Error rates are calculated as a percentage
of records or transactions.
Quality and Automated Reporting
The increased use of automated reporting has generally improved the quality
of the merchandise trade statistics. The Census Bureau now receives more
than ninety-five percent of import transactions and more than fifty-five
percent of export transactions electronically. As explained earlier, automated
reporting allows the Census Bureau to receive and compile the data quickly
and include almost all shipments in the correct statistical month. Automated
reporting minimizes lost data and provides better control over non-filing.
Finally, since Census receives pre-edited data through the U.S./Canada
Data Exchange and ABI as well as partially edited information through the
AERP, the information arrives at the Census with fewer reporting errors.
Recent studies show that about fifty percent or one in two export records
captured from SEDs (export paper documents) contains an error as compared
to about eight percent or one in eleven AERP or Canadian data exchange
records. Likewise, about one in four data records captured from Customs
Entry Forms 7501 (import paper documents) contains an error as compared
to one in twelve ABI records.
The enhanced quality of automated submissions results from the high standards
required of automated reporters. The Customs Service and the Census Bureau
test and approve all ABI brokers and AERP participants before they can
submit shipments electronically. Many statistical Census edits reside in
the Customs computer system and intercept data problems for return to the
ABI broker for resolution. ABI participants who fail to resolve problems
or who repeatedly make the same errors face possible fines or penalties.
AERP participants also must maintain acceptable levels of performance or
face removal from the program.
Similarly many Census edits reside in the Statistics Canada import processing
system to meet U.S. export needs. Also the statistical and customs agencies
of each country have designated staff that meet regularly to resolve any
problems and keep abreast of any changes that may affect the export programs.
To reduce further export filing on SEDs, the Customs Service is working
on the development of an Automated Export System (AES) similar to the ABI
system. AES will link to other Customs automated tracking systems to help
prevent undocumented shipments. It also will incorporate Census statistical
edits to improve the quality of reported data.
The Customs Service conducted a sixty day test of AES using export transactions
March and April 1993 at the port of Charleston, South Carolina. Participating
carriers transmitted manifest data directly to Customs. Corresponding statistical
merchandise trade data were keyed from SEDs and transmitted to Customs
by the Charleston port authority.
Initially the AES edits intercepted statistical reporting errors on over
forty percent of the SEDs transmitted to Customs. The reasons for rejection
involved missing or incorrect quantities or units of measure (68 percent)
and invalid HS classification codes (32 percent). This high percentage
of reporting errors fell to three percent after participants adopted special
screening and editing processes used by the Census Bureau to identify and
capture statistical data from incorrectly prepared SEDs.
The AES test proved that the electronic interface between the trade and
Customs can receive and transmit data and that the edits and validations
planned for AES will significantly improve the quality of export statistics.
Despite these accomplishments, the Customs evaluation team found significant
flaws in the conduct of the test and recommended more testing before national
implementation.
The Census Bureau continues to support strongly AES. However, until AES
is operational the Census Bureau will continue to expand the AERP program.
III. Adjustments
The Census Bureau adjusts merchandise trade data to remove the effects
of seasonal influences and price shifts. Adjusted data appear in the monthly
FT900 "Advance Report on Merchandise Trade" released by the Department
of Commerce. The adjustments remove known or predictable influences that
may obscure true changes in the volume of merchandise traded.
Adjustment for Seasonal and Trading Day Variation
These adjustments, computed using the Census Bureau's X-12-Arima program
(a variation of X-11 with Arima modeling), remove predictable calendar
effects from the data. X-12 develops seasonal factors from historical series
of data. The merchandise trade series exhibit a high level of irregular
variation. To help overcome this problem, the Census Bureau uses the mathematical
modeling of X-12 to adjust for level shifts and outliers in the data series.
The choice of model can affect the calculation of factors and will change
over time as more data are added to the series. Generally the use of longer
series will help clarify the seasonal patterns.
However, the adjustments will not remove all seasonal variation. If the
timing and magnitude of the seasonal effect is not predictable or stable
the Census Bureau cannot adjust the series. As working day effects (the
differences between the level of trade on specific days of the week) are
generally small, the high irregular variation of many series prevents accurate
identification of these movements. Thus, while the final published series
show no clear evidence of working day effect, some residual influence may
be present.
Adjustment for Price Change
Besides seasonally adjusting the data, the Census Bureau also publishes
trade data on a constant dollar basis. The adjusted data provide an estimate
of the change in the volume of trade, as distinct from value changes resulting
from price shifts. The Census Bureau bases the adjustments on the International
Price Indexes (IPI) published by the Bureau of Labor Statistics (BLS),
supplemented with deflators produced by the Bureau of Economic Analysis
(BEA).
Only a few indexes are available for the first two months of each quarter.
BLS makes a complete set of end use indexes available for the final month
of each quarter. Using a procedure developed by BEA, the Census Bureau
extrapolates the detailed indexes based upon the higher level monthly indexes.
As a result, changes from the second to the third month of each quarter
may be larger than for other months. Studies show that these procedures
do not distort the indexes. However, month-to-month changes in the detailed
data should be used with caution.
The Census Bureau deflates trade in computers and peripherals using the
quality adjusted hedonic deflator computed by BEA. BEA uses the attributes
(speed, memory capacity, etc.) of the computer equipment to estimate the
value of quality improvements over time. Prices relative to quality have
dropped rapidly for this product grouping. As a result, computers and peripheral
equipment comprise a larger part of total value on a constant dollar basis
than on a current dollar basis. Users may wish to examine this area of
trade separately from total trade.
III. Conclusion
The merchandise trade statistics program was originally designed to
meet the basic needs of policy makers and provides reasonable estimates
of total trade with other countries. However, the program has grown in
scope and complexity, attempting to meet the needs of many different data
user groups. For example, the program collects transportation data for
air and vessel carriers, port authorities and the Department of Transportation.
It also collects detailed commodity information for trade associations,
marketing analysts and government agencies investigating the impact of
international trade on local industry and economies.
The needs of data users expand and change continuously. The Census Bureau
and the Customs Service strive to provide accurate and complete information
for all data users. Realistically, however, the program meets the needs
of some data users better than others.
The Census Bureau, to address the concerns of these many data users, will
continue to release annual quality profiles. These profiles will include
the results of studies conducted to assess the quality of the data and
any reports on the results of quality assurance programs. The profiles
also will describe any changes in the programs that may effect the quality
of the statistics. In this way, the Census Bureau hopes to provide the
information and knowledge needed to make it's many data users "educated
consumers" of the merchandise trade data.
Appendix A
Annotated Bibliography
Bureau of the Census, 1992, "Adjustment of Foreign Trade Data for Seasonality
and Price Change." Foreign Trade Division, Bureau of the Census, U.S. Department
of Commerce, Washington, DC.
This technical paper explains the origin of the constant dollar series
and the methods used to seasonally adjust and deflate the merchandise trade
data. Contact: Diane Oberg (301) 763-5708.
Bureau of the Census, 1987, "U.S. Merchandise Trade: July 1987," U.S. Department
of Commerce News. FT900 CB-87-147. Washington, DC: U.S. Department of Commerce.
This issue contains the special notice on improved data coverage with Canada
and gives undocumented exports annually for 1970 through 1988 and monthly
for 1987. Contact: Trade Information Office (301) 763-5140.
Bureau of the Census, 1992, "U.S. Merchandise Trade: December 1992," U.S.
Department of Commerce News. FT900 CB-93-36. Washington, DC: U.S. Department
of Commerce.
This issue contains the special announcement concerning imports of motor
vehicles from Japan not reflected in the import trade statistics. For more
information, contact Paul Herrick (301) 763-5201.
National Research Council, Panel on Foreign Trade Statistics, 1992, Behind
the Numbers: U.S. Trade in the world Economy. Washington, DC: National
Academy Press.
This report covers issues related to the merchandise trade statistics and
the need for information to guide the public and private sectors in the
current world economy. From a quality standpoint, it focuses on improving
data collection, accuracy and analysis.
