Chapter 2 - Method for Creating Trade-flow Map/Trade Matrix

A trade matrix is a numerical table showing data on trade between region blocks and between major countries, in matrix format. All numerical data in the trade matrices stored on the CD has been converted into weight units for uniformity. A trade-flow map shows world trade by means of arrows on a map, and is created from trade matrices. This chapter describes the method for creating trade-flow maps and trade matrices.

1. The Data

Every year, the United Nations tabulates trade statistics from the trade data reported by various countries, and offers these statistics on magnetic media (henceforth, the “United Nations Trade Statistics Dataset”). In addition, the International Trade Center (ITC) offers a Personal Computer Trade Analysis System (PC-TAS), which is derived from and processes UN Trade Datasets since 1995. Data has been compiled for 4 years: 1983, 1988, 1993, and 1998.

Some of the original trade data is only reported in monetary units, and different countries use different units of quantity measurement for their trade data. For this reason, when creating a trade matrix, it was not possible to simply tabulate the data for a given region; it was also necessary to convert all transactions into units of weight. The statistical method for doing this is described in detail in section 3. Below are described the salient features of the data.

(1) Data Format

For each type of trade, the UN Trade Dataset contains data on trade-flow type (import, export, or re-export), commodity, reporting country, partner country, transaction amount, unit of quantity measurement, quantity traded, and summation period. As there are both import and export reports, if every country in the world correctly reports its trading, then all trade data will be contained in the dataset twice, once as import data and once as export data.

(2) Level of Coverage of Total Trade

There are currently about 100 reporting countries (94 in 1983, 84 in 1988, 91 in 1993, and 100 in 1998). Since the developed countries, which consume the majority of resources, generally report their trade, it is thought that most trading by non-reporting countries is accounted for as well via their partners. It must nevertheless be noted that trade between non-reporting countries is completely missed.

(3) Commodity Categories and Category Codes

Commodities are classified according to the UN’s Standard International Trade Classification (SITC). There are 3 versions of the SITC: Rev. 1 (about 1,800 categories), Rev. 2 (about 2,600), and Rev. 3 (about 4,200). With each revision the classifications become finer, providing a more detailed breakdown of commodities, while at the same time with each revision, there are fewer reporting countries. Thus, in order to strike a balance between granularity of classifications and number of reporting countries, the data from Rev. 2 was used. Each commodity has a commodity code of up to 5 digits, with each additional digit providing a finer-grained classification. For example: 0:FOOD AND LIVE ANIMAL / 01:MEAT AND PREPARATIONS / 011:MEAT FRESH,CHILLD,FROZEN.

(4) Level of Coverage of Quantity Data

Unlike monetary data, quantity data for trade sometimes contains empty (blank) fields. In some cases, all fields for a given commodity will be blank, while in others only some of them will be blank. Trade quantity is measured using many different units, including weight, volume, area, and count. Quantities of different commodities are reported using different units, and in some cases different countries may report the same commodity using different units.

2. Commodities Covered

As shown below, the trade matrix contains resources taken from nature, and raw materials with low levels of processing, which have relatively high trade flows, and which are considered vital from an environmental standpoint. It excludes resources with low trade flows, such as precious metals, and commodities with high levels of processing, like machinery and chemicals. Lime, gravel, and other non-metal minerals were also excluded.

· Food resources (meat, fish and grains)

· Wood resources (wood and products made from wood with little processing)

· Metal resources (ore and metal products)

· Fossil fuel resources (including converted fuels)

Fig.1 is a trade matrix showing the hierarchical relationships of these resources.

3. Creating the Trade Matrix

Fig.2 shows the trade-matrix compilation process, while Table 1(.xls) lists the commodities included in the matrix. The trade matrix covers 50 commodities over 3 years, and thus contains 150 items. Note that in cases where commodities covered by the matrix corresponded to commodities in Japanese Ministry of Finance (MOF) trade statistics, it was confirmed that the data on Japanese imports (the section corresponding to column 9 of the trade matrix) matched the MOF trade statistics nearly exactly. Below is a description of the process used to create the trade matrix.

(1) Convert 1998 Trade Data (PC-TAS)

After extracting necessary 1998 trade data from PC-TAS, it was converted to United Nations Trade Statistics Dataset record format. Additionally, as PC-TAS uses the SITC Rev. 3 commodity codes, a correspondence table provided by the United Nations was used to convert the codes to SITC Rev. 2.

(2) Extract Corresponding Commodity Records (create Workfile 1)

Records matching the commodities described in 2 were extracted from a set of several million trade-data records.

(3) Adjust Import/Export Data (create Workfile 3)

As described in 1-(1), the UN Trade Dataset consists of import data and export data; if both are used, trade between reporting countries will be double counted. It is therefore necessary to use one or the other. For the present tabulation, it was decided to use the import data, and discard the export data. This discarded export data, however, includes data on trade with non-reporting countries, which is not contained in the import data. Thus, data on exports to non-reporting countries was converted to import data, and used in lieu thereof. Additionally, in the case of trade with Japan, the values reported by Japan to the UN are used in all cases; namely, out of the values extracted from the import data of the UN Trade Dataset, the records showing imports from Japan were removed, and replaced by Japan’s export values to the country contained in the discarded export data. It further must be noted that trade between non-reporting countries is completed unaccounted for.

(4) Statistically Extrapolate Quantity Data (create Workfile 3)

As described in 1-(4), the transaction quantity fields of some of the records in the United Nations Trade Statistics Dataset are blank, and in some cases units of quantity measurement other than weight are used. For these records, it is necessary to statistically extrapolate the transaction quantities in units of weight. Before conducting this extrapolation, the reporting of quantity units for each commodity was reviewed; this review showed that weights were reported in a large number of cases for all commodities except “Saw-, Veneer-logs” for 1988. Thus, the following method was used to statistically extrapolate the weight units for all records except “Saw-, Veneer-logs” for 1988:

・ First, records for which weights were reported were extracted, and transaction amounts and weights tabulated. This information was used to calculate an average price per unit of weight.

・ The average price per unit weight was then used as the divisor in dividing transaction amount of records where a non-weight unit of measurement quantity was used, or where the quantity field was blank, in order to obtain estimated transaction weight.

Meanwhile, for “Saw-, Veneer-logs,” about the same number of records had quantities in terms of volume as weight. Believing that it would be inappropriate to discard one portion of the data and use the other, it was decided convert volumes to weights by applying a conversion coefficient obtained from another source, and then apply the method described above.

During the course of the check, the unit price of each record for a given commodity was calculated. When these calculations were compared, it was found that in a small number of records, the prices were off by several orders of magnitude from the other records. Although it was believed that in most cases, the discrepancy was due to simple input error during reporting, it was hard to decide which cases were actually mistakes. Thus, it was decided to compare the price of each record with the median value, and if there was a difference of 3 digits or greater, to treat the quantity for that transaction as an empty field.

(5) Convert to Matrix Format

Commodity codes with few digits, i.e. high-level commodities, represent several different lower-level commodities as a single category. Thus a high-level commodity may subsume items with substantially different unit prices, due to different levels of processing and different qualities. The statistical method used here, however, converts all monetary amounts for given commodity to weights using the same unit price, which in the case of these commodities would cause the calculated weights to be too high for expensive commodities, and too low for inexpensive commodities, significantly degrading the accuracy of the statistical extrapolation. In order to avoid this problem, the method described in (4) was not used for high-level commodities. Instead, it was necessary to sum up the data created for lower-level commodities.

Additionally, it was necessary to create a trade matrix of commodities not contained in the SITC, as in the case of biomass where a new trade matrix was created by combining the food resource and wood resource trade matrices (see Fig.1).

Trade matrices were first created for commodities relatively low on the hierarchy, as “base” commodities. Next, these trade matrices were summed, in order to create trade matrices for higher-level commodities, and commodities not in the SITC. If, however, there are cases where only higher-level commodities are reported, and lower-level commodities are not reported, then there is a chance that some trade will be missed. For this reason, the monetary amounts of commodities summed for estimation purposes were compared with the monetary amounts of the higher-level commodities, in order to make sure that there were no problems with this process.

Table 2(.xls) shows the correspondence relationship when totaling and recombining commodities. The commodity codes for trade matrices created via this summation process are determined as follows:

・ Commodity codes starting with “T” indicate that the commodity has been uniquely defined here, due to the fact that it represents a broad category of major resources.

・ Commodity codes ending with “A” exist in the SITC except for the “A,” but indicate that the commodity has been recombined into a different makeup.

・ Commodity codes ending in “0” indicate that the commodity was created by summing lower-level commodities.

a) Creation of Base-commodity Trade Matrices

Base-commodity trade matrices were created from Workfile 3 using a trade-matrix creation program (a program that creates trade matrices from data in OD pair-record format, and outputs them in spreadsheet format) developed by the National Institute for Environmental Studies. The commodities in the trade matrices created by this program include commodities not containing a “T”, “A”, or “0” in Table 1. Tables 4(.xls) and 5(.xls), respectively, list the names of the region block classifications (Country/Region Block Tabulation Codes), and the names of each country in each region, used for the tabulation.

b) Commodity Summation and Recombination

A commercial spreadsheet application was used to sum and recombine the trade matrices created in a).

4. Creating the Trade-flow Map

A trade-flow map was created using Acclaim, a system developed by the National Institute for Environmental Studies that shows the international trade balance of environmental resources. Table 6(.xls) lists the items contained in the map.

5. Notes on Interpreting the Trade Matrix and Trade-flow Map

Below are notes on interpreting the trade matrix and trade-flow map.

(1) The Trade Matrix

・ The rows of the matrix are the exporters, and the columns the importers.

・ An ID code is assigned to each trade matrix. The ID has the format YY-XXXXX, where YY is the two-digit data year, and XXXXX is the (up to 5-digit) commodity code.

・ It must be noted that this data has been statistically derived from the United Nations Trade Statistics Dataset.

(2) The Trade-flow Map

・ The arrow flows show the net trade quantities (balances) between two countries/regions.

・ Due to the nature of the trade-flow map, flows within a single region (e.g. between European countries) are not shown. Refer to the trade matrices for this information.

・ Due to the nature of the trade-flow map, trade with Other East/South Asia is not shown. Refer to the trade matrices for this information.

・ For ease of viewing, the Asian Region version does not show trade between non-Asian region blocks.

(3) Re-export

・ The UN Trade Datasets from 1983 to 1993 carried trade data on re-exports. Here, the data on re-exports was counted as export data.

・ Note that the 1998 PC-TAS does not deal with data on re-exports.

Chapter 2 - Method for Creating Trade-flow Map/Trade Matrix

Tables and Figures