3. Data Base Processing
- International Monetary Fund
- Published Date:
- March 1993
The collected reported data are then processed in the data base for the direction of trade statistics. In turn, the availability of current trade data, supplemented with estimates for individual countries, facilitates the continuous data base update of world and area table aggregates. Section 3.1 explains how the data base reformats reported data to impart uniformity before data are stored, section 3.2 outlines estimation procedures, and section 3.3 reviews data base compilation of aggregates for world and area tables.
3.1 Reformatting of Reported Data
DOTS source data are processed in the data base for direction of trade statistics on an ongoing basis as data are received; updates to the master file are normally performed every week. After Statistics Department staff review reported data for conceptual validity, the data base performs computerized quantitative and editorial checks (“horizontal” and “vertical”) and validation tests. First, the reported total of exports or imports is compared against (1) the total calculated from the detailed data by partner countries and (2) the data on total exports or imports reported independently for publication in IFS. Secondly, the data by an individual partner country are compared for validity against the observations for previous periods of the same year and the same period of previous years.1 After the reported data have passed the validation checks,2 the data are modified before storage to conform with common characteristics of the data base—namely, noncumulative monthly, quarterly, or annual frequency, U.S. dollars unit, exports f.o.b. and imports c.i.f. valuation, and a specific classification of countries (the IMF country and partner country codes).
The data base converts individual country data, usually reported in national currencies, into U.S. dollars. All data in the master file for direction of trade statistics are stored in U.S. dollars as the common unit of account in order to ensure international comparability. To convert data, the data base uses exchange rate series rf published in IFS;3 series rf refers to period averages of market exchange rates (or official rates if market rates are not available). Data are converted at their highest available frequency and subsequently aggregated to longer periods; for example, monthly data are converted to U.S. dollars and then aggregated to quarterly and annual totals. The data are then stored in the master file in millions of U.S. dollars, with one to three decimals, depending on the country.
3.1.2 C.i.f. and f.o.b. valuation
The majority of reporting countries follow the United Nations’ recommendation to record exports on an f.o.b. basis and imports on a c.i.f. basis. A few countries, however, compile and publish imports on an f.o.b. basis,4 and their data are shown on this basis in the country pages of the DOTS quarterly issues and yearbook. In addition, these import data reported f.o.b. are stored in the master file in three categories (see also section 5.1): (1) imports f.o.b. (for publication purposes), (2) imports c.i.f. adjusted on the basis of a 10 percent factor (stored for estimation purposes),5 and (3) imports c.i.f. adjusted on the basis of the factor used for calculating the world import table in IFS. (For the calculation of world and area tables, see section 3.3.)
3.1.3 Partner country codification
The DOTS source data are reported under differing partner country classifications (country names or codes). While the majority of countries use country names, several Latin American countries use codes from their own system. European Economic Community members follow a common country classification.
All reported partner country names or codes are automatically converted into the IMF partner country code (see Appendix I), with the help of the so-called dictionary—a special file containing all possible country names and codes as used by the reporting countries. This file is comprehensive; however, problems can occur if a country does not provide an exhaustive list of trading partners.
Special IMF codes are used for “not specified” categories. Some countries report area subtotals in addition to individual country data. When the sum of data for the listed countries does not match the corresponding subgroup total, a residual or area total “not specified” is created by subtraction (for example, “Africa not specified”). More often, the reporting countries themselves report the category “not specified.”
3.1.4 Summation and area totals
A country’s exports to and imports from countries within areas are aggregated and stored as area totals. Currently stored in the direction of trade statistics data base are 11 area aggregates, in addition to the world total, which is obtained by summing the area totals.
Fund Statistics Department staff calculate area and world totals according to the country classification in IFS. The classification distinguishes three main categories: industrial countries, developing countries, and a third group of countries, summarized as other countries n.i.e. (not included elsewhere). The second category, developing countries, is divided into five areas: Africa, Asia, Europe, Middle East, and Western Hemisphere. In the absence of a more suitable term that would conveniently cover the countries included in the third category, they are referred to as other countries n.i.e. This third category comprises Albania, Bulgaria, Cuba, Eastern Germany,6 Mongolian Republic, North Korea, and the former U.S.S.R., which are not included in the world trade table published in IFS. The country composition of each area of the world is listed in Appendix I.
Totals for the subgroups oil exporting countries, non-oil developing countries, and the European Economic Community are also calculated as memorandum items. The country composition of these subgroups is given in Appendix I.
Each country reports its trade data to differing degrees of currentness, frequency, and detail. As explained in section 2.3, DOTS correspondents in approximately 40 countries report monthly trade data by partner country on a regular and current basis. These countries, comprising virtually all the industrial countries and about 20 developing countries, have represented in recent years nearly four fifths of the value of recorded world exports and imports. Some other countries may report less current monthly data, or they may compile and make that data available quarterly or annually. Other countries may report data only on total exports and total imports, with no breakdown by partner country. Finally, for other countries, no current trade data may be available.
The data management system for the direction of statistics data base makes it possible to provide supplementary estimates for individual countries, whenever reported data are not current or available in a monthly frequency. Therefore, monthly data for all countries are available with a delay of four or fewer months from the current month.7 The estimation procedure is tailored for each country depending on the currentness, frequency, and degree of detail of trade data that are reported to the Fund and that are therefore available as benchmark data for estimates.
3.2.1 Estimation procedure
This section provides a general description of the estimation procedure. (A more detailed description of the compilation of estimates is given in Appendix III.)
Procedure for generating estimates of a country’s trade
The data management system for the direction of trade statistics data base generates estimates of a country’s trade with each of its partner countries, for any month for which that country has not reported data for inclusion in DOTS. (Subsequently, when a country reports its own data, then the estimates for its trade with all its partners are removed.) The basis (benchmark) for all estimates is a country’s reported data, stored in the Fund’s data base, including total exports and imports reported independently and published in IFS. The benchmark data used for the estimates are, first, lower frequency (quarterly and annual) trade data by partner country, if available, from the country’s own sources; then, monthly data on total exports and imports reported for IFS, if available, and finally, a combination of reported and estimated partner country data, and extrapolated data.
The matrix nature of the Fund’s data base for direction of trade statistics facilitates the filling of data cells with data reported by trade partners; recourse to estimation by extrapolation thus is limited to very small sections of world trade. Moreover, data from countries’ own records, even with longer delays, continually broaden the base of reported data, thereby eliminating previously estimated figures.
Procedure for deriving estimates from partner country data
In those cases where partner data are used to derive estimates, such data are first adjusted by a uniformly applied percentage to allow for the cost of freight and insurance, assumed to be 10 percent of the f.o.b. value of imports. For example, if estimates for country B are based on data reported by country A, then A’s data for imports c.i.f. from B, reduced by 10 percent, are taken to be the f.o.b. value of B’s exports to A; conversely, A’s data for exports to B, augmented by 10 percent, are taken to be the c.i.f. value of B’s imports from A. The allocation of 10 percent is a highly simplified estimate used as an overall approximation to the actual cost of freight and insurance, which may vary widely from country to country.8
Procedure for extrapolation
Extrapolation is used in those cases where the derivation procedure cannot be applied, because neither the exporting nor the importing country has reported or estimated data flows for the period in question. In those instances, trade flows are estimated by extrapolating data, reported or estimated as available, for the same month of the previous year. Extrapolation factors are based on an assessment of trade developments in large areas of the world (indicated in available reported data) and are regularly updated and revised.
Breaksigns for estimates
Breaksigns indicate monthly and consolidated quarterly and annual estimates for individual countries in the data base (see Appendix II for breaksigns). These breaksigns refer only to figures for trade between two partner countries and are not applicable to totals calculated for groups of partner countries, nor are such symbols found in figures for world and area tables.
Timeliness of estimates
To keep data up-to-date in the DOTS data base, with a delay of four months from the current month, estimates are updated each month. They are also revised each time reported data are updated in the master file (normally every week).
3.2.2 Limitations in DOTS estimates
Caution should be exercised when using direction of trade statistics estimates, because estimation procedures for each month vary to the extent that data are available for each country. Furthermore, estimation methods are based on assumptions about trade patterns during previously reported periods, about change in trade for reporting countries, and about comparability of international trade flows between partner countries. The following cases of estimation illustrate assumptions about trade:
Limitation in distribution of reported DOTS data
When the data management system for the direction of trade statistics data base automatically distributes lower frequency (quarterly and annual) DOTS reported data into months, it uses the monthly pattern displayed by monthly reported partner data, where available. If those data are not available, it bases the monthly pattern for the remaining partner countries on the reported monthly total exports or imports for IFS. If IFS monthly data are not available, then data are evenly distributed across months. Therefore, although monthly distributed data are constrained to adding up to reported quarterly or annual reported data, the estimated monthly pattern may differ from the actual data.
Limitation in distribution of total IFS exports and imports
When DOTS estimates for a country are benchmarked on the monthly total exports and imports reported by the country for IFS from the country’s own sources (with no breakdown by partner country), the data base management system calculates the partner country pattern on the basis of the most recent annual DOTS trade-by-country statistics reported by the country. This method is equivalent to extrapolating export and import figures by partner countries on the basis of the monthly total exports or imports reported by the country. The implicit assumption for the estimates is that the country’s geographical trade patterns did not change from previously reported periods. However, in practice, geographical trade patterns may change during short periods, owing to the unstable nature of trade flows.
Limitation in estimates derived from partner country data
When partner data are used to derive estimates, the method presumes the comparability of trade flows between trading partners. In practice, however, figures for trade flows between partner countries differ for several reasons, like unavoidable valuation and time lag differences and methodological differences between national systems for recording trade statistics.9 Regarding the c.i.f./f.o.b. valuation adjustment, the DOTS estimation procedure applies a uniform percentage, assumed to be 10 percent of the f.o.b. value of imports, as a highly simplified approximation of the actual costs of insurance and freight. As for time lags, no adjustment is made, because time lags vary widely between bilateral trade flows and depend on such factors as distance between partner countries and means of transportation. However, differences in timing are unavoidable, particularly for data compiled on a monthly basis.
Limitation in estimates based on extrapolation
When the data management system extrapolates trade flows between two countries, it multiplies the data for the same month of the previous year—reported or estimated as available—by a factor that takes into account trade development in the areas in question. Herein, a limitation in estimates based on extrapolation is illustrated: The implicit assumption is that changes in trade flows between two countries can be represented by changes in trade flows between all countries of the areas in question. Another limitation in this method is that because estimates can be extrapolated over several years the errors may be cumulative.
3.3 Compilation of Aggregates for DOTS World and Area Tables
Once reported trade data for individual countries have been formatted as described in section 3.1 (and supplemented, if necessary, by estimates), they are compiled by the data base system into world and area aggregates, which show flows between an individual country and areas of the world, between an area and individual countries, and between areas. Whereas, before the data base was upgraded, these aggregates had been available only yearly (in the DOTS Yearbook world and area tables), they are now also available monthly in the data base, updated and revised at the end of each production cycle for the monthly DOTS computer tape.
World and area aggregates are calculated in the data base in two ways, called “Part A” and “Part B,” as published in detail in the DOTS Yearbook. Aggregates for Part A present trade flows of areas as a whole with individual countries, as reported and estimated for all countries in that area. Aggregates for Part B present trade flows of individual countries with areas on the basis of the world and area lines shown on the country pages.
Part A shows the country and area distribution of aggregate trade of an area or the world with countries listed in the table for that particular area. This can be visualized by imagining all pages for individual countries in an area placed in a stack and all figures added vertically through to an aggregate table on the top of the stack; the resulting table would be the Part A table for that area.
Part B, the second set of world and area tables, presents the trade of the countries listed in the table with the area covered in that table. The lines for each country in Part B tables are a transcription of the area and world lines for the individual countries, and the area and world lines in Part B are the sums of these transcribed lines. The Part A summary tables focus on the areas’ trade with individual countries; the Part B summary tables focus on individual countries’ trade with areas.
In compiling these tables, the data base system adjusts imports, reported f.o.b. by individual countries, to c.i.f. values by application of the factor used for calculating these countries’ imports c.i.f. for the world imports table and the country pages in IFS.10
A comparison of Parts A and B indicates the degree of symmetry and accuracy in the recording of trade flows. In the absence of problems of valuation, timing, and coverage, the exports in Part A, plus freight and insurance, should approximate the imports in Part B and vice versa. However, as discussed in section 1.4.7, a number of factors give rise to differences between these trade flows.
For data processing (e.g., decumulation), validation checks, and archive purposes, a so-called reported data file is maintained. This file contains the data in the same format (with no transformation) as reported by countries.
When data fail the validation test, the data are compared against the source document and alternative publication sources, and judgment is used to accept the data or contact the national DOTS correspondent.
See Kar, “Currency Invoicing and Exchange Conversion in International Trade,” op. cit.
A few additional countries publish imports on another basis (for example, “valued at site” or customs value basis).
The 10 percent factor is a rough approximation of the value of the costs of insurance and freight. This factor is also used in cases where estimates are derived on the basis of partner data (see section 3.2).
The data for Eastern Germany include figures through June 1990 and cover the area of the former German Democratic Republic.
No estimates are compiled for periods prior to 1981 or based on benchmark data referring to 1980 or earlier.
While ideally there should be a different factor for each individual trade transaction for each time period, the 10 percent factor is a rough approximation of the valuation adjustment. The use of a constant factor has the additional advantage of being simple and manageable. However, the uniform factor has limitations because it may not apply strictly to all bilateral flows.
A more complete discussion on the comparability of trade flows between trading partners is given in section 1.4.7.
See also the c.i.f./f.o.b. factor table in the IFS yearbooks.