Economic Commission for Africa

I ** ^ ECA/DfSD/4.2cl/2(v)/1999 Economic Commission for Africa Methodological issues in the compilation of international trade indices in African countries: Case study of selected African countries Addis Ababa, Ethiopia September 1999

ECA/DISD/4.2d/2(v)/1999 Table of contents Paragraph I. Introduction 1~3 I1. Objectives of the study 4-5 HE. Case studies of selected African countries 6-130 A. Kenya - Part I 7-36 (i) Introduction 7-12 (ii) Detailed examination of selected commodities 13-22 (iii) Methodology for compiling trade indices 23-24 (iv) Splicing and linking/chaining 25 (v) Examination of specific problems 26-30 (vi) General problems 31-35 (vii) Summary 36 A. Kenya-Part II 37-51 (i) Introduction 37-38 (ii) Selection of items 39 (iii) Sample coverage for imports 40-42 (iv) Sample coverage for exports 43 (v) Assignment of weights 44 (vi) Inclusion or exclusion of re-exports data In the export trade indices 45-48 (vii) Revisit the decision of making the 1994 as base Base year of the trade indices 49-50 (viii) Technical calculation of index numbers 51 B. South Africa 52-81 (i) Introduction 52-57 (ii) Processing and dissemination 58-60 (iii) Compilation of foreign trade indices at Statistics South Africa 61-65 (iv) Compilation of trade indices at Statistics South Africa 66-71 (v) Other organisations involved in trade statistics 72-75 (vi) Recommendations 81 C. Mauritius 82-131 (i) Introduction 82-83 (ii) External trade statistics-general aspects 84-90

ECA/DISD/4.2d/2(v)/1999 (iii) Foreign trade unit value indices 91-97 (iv) Calculation 98-102 (v) Domestic exports 103-109 (vi) Export processing zone 110-119 (vii) Export price index 120-131 IV. Conclusions and recommendations 132-136

ECA/DISD/4.2d/2(v)/1999 I. Introduction 1. The Development Information Services Division (DISD) of the Economic Commission for Africa (EGA) work programme for the biennium 1998/99 included the preparation of the non-recurrent publication titled "Methodological issues in the compilation of international trade indices in african countries: case studies of selected African countries". This publication was intended to be one of the technical publications of ECA. Also to be an input into the documentation of an Ad-hoc expert group meeting on improvement of intra-african trade statistics with emphasis on the use of statistical software packages, planned to be organised by ECA during the second half of 1999. 2. Three countries: Kenya, South Africa and Mauritius were chosen for the case study. To enable the compilation of the relevant information for the study, ECA undertook missions to those countries as follows: Kenya (14-15 November 1996 and 1-5 December 1997); Mauritius (16-20 November 1998); and South Africa (23-27 November 1998). 3. As will be noted, the two missions to Kenya were undertaken before the study was included in the 1998/99 work programme of the ECA. The reason for this was that the Central Bureau of Statistics (CBS), Kenya, was in the process of rebasing its international trade indices and as such it requested ECA assistance. The CBS indicated that its basket of goods for the calculation of trade indices was outdated hence it did not reflect the true picture of changes that were taking place in internationally traded commodities. At the same time the CBS was reveiwing its methodologies for the compilation of various deflators. Due to the nature of the problems, which were experienced in the past, in doing this type of work, the CBS felt that there was need to address a number of methodological issues with a view to improving the overall quality of the trade indices. 11. Objectives of the study 4. The objectives of the study were two fold: (a) to present methodologies which were adopted by the three countries in the compilation of the international trade indices; and (b) to evaluate those methodologies and make recommendations. 5. The hope was that this methodological study would assist many African countries in their work on compiling international trade indices, as an a nalysis tool, to enable them to meet the data needs of many users in this field. At the time of writing this report, the majority of African countries were not compiling international trade indices, it was therefore the hope that this technical study would encourage as many African countries as possible to compile trade indices. Ml. Case studies of selected African countries 6. Three case studies are reported: The case study of Kenya has the theme "Rebasing trade indices". The case study of South Africa has the theme "Towards reengineering the compilation of trade indices"; and the case study of Mauritius has the theme "Fully developed set of trade indices".

ECA/DISD/4.2d/2(v)/1999 Page 2 A. Kenya - Part I Re-basing trade indices (i) Introduction 7. The initial study was undertaken during a mission to the Kenya Central Bureau of Statistics from 14 to 15 November 1996. The background to the Kenya trade indices is described in Annex I, which is an extract of an article published by the CBS in the Kenya Statistical Digest of June 1977. Based on that background, the CBS undertook an initial analysis of the situation on import/export trade indices. This analysis enabled the CBS to understand the methodological issues and to see to what extent those issues could be addressed during the process of rebasing the trade indices. 8. The trade indices, which were being compiled by the CBS had 1982 as the base period and as such they did not reflect the true picture of changes that had taken place in internationally traded commodities, trade indices. The current basket of goods was outdated and as such it did not reflect the true picture of changes that have taken place in internationally traded commodities. Due to the nature of problems experienced in the past in doing this work, the CBS felt that there was need to address a number of technical issues with a view to improving the quality of the indices. 9. The trade indices at that time included 310 commodities based on SITC Rev.3 classification. The new index was proposed to have a 1994 base period. The choice of this year was influenced by the fact that the external sector of Kenya was liberalised during 1994. 10. The customs data was proposed for use in calculating the indices due to the fact that it is readily available, hence no extra resources would be needed to collect data as would be the case if an import/export survey was to be conducted to collect price data directly. Data on SITC Section 9 would not be included in the index due to lack of specification of items or due to the nature of items included in that SITC section. The Laspeyres, Paasche and Fisher index formulae would be used in the calculation of the rebased trade indices. 11. A number of problems which could affect the new index numbers were discussed. Among them the quality of imported commodities which varied from country to country; some of the commodities were subsidised while others were not; some importers undervalued imports to avoid high import duties; and there was heterogeneity of imported/exported products. The quantities of some commodities were at times not correct and in some cases, such as in chemicals, there could be lack of specification particularly on quantities. Further, goods in transit which found its way into the country tended to distort prices on the market. These goods were to some extent captured by the consumer price index. The problems of SITC section 7 commodities, particularly transport and machinery were mentioned. This problem was more on the heterogeneity of commodities particularly in the electronics field.

ECA/DISD/4.2d/(v)/1999 Page 3 12. The initial idea was to examine data covering three years 1993, 1994 and 1995. The unit prices would form the basic variable in the examination process. (ii) Detailed examination of the selected commodities 13. Detailed discussions were held with the staff of the Trade, Finance and Balance of Payment Section of the CBS who are engaged in the work of compiling trade indices. 14. Two criteria were chosen by the CBS for the selection of commodities to be included in the computation of the indices. In the case of exports the criteria was: (a) select a commodity which has been exported to at least five (5) countries; and (b) the value of exported commodities should not to be less than 20,000 Kshillings. In the case of imports the criteria was: (a) select a commodity which has been imported from at least five (5) countries; and (b) the value of the imported commodity should be at least I million Kshillings. 15. It was clear from the beginning that there was need to make a general recommendation on problems of outliers in the initial selection of commodities. For each selected commodity, on the basis of the calculated unit values, outliers should be identified and be removed. A new unit value should be calculated. In doing this, the requirement for number of countries from which or to which the commodity has been imported/exported should be fulfilled if the commodity is to remain in the sample. Otherwise if the number of countries falls below 5 then the commodity should automatically be dropped from the sample. 16. SITC section 0: Food and live animals: In examining a selection of commodities, it was clear that maize should be split into seed maize and maize for food. It was observed that by removing outliers, the variation in unit price in time-series would be resolved for the majority of commodities in the group. As for subsidies which may be applicable to certain food commodities, it was felt that this would be taken care of by market forces in the economic management of the Kenya economy. It was not clear how to deal with live animals since some may have been imported for the Zoo or for sport, etc. It was however recommended to exclude animals destined for the Zoo. 17. SiTC section 1: Beverages and tobacco. The problem which was identified was that of consumer preference and quality of products. By removing outliers, it was felt that unit values for the selected commodities would be stabilized. 18. SITC section 2: Crude materials, inedible, except fuels. It was recommended that items labelled "other" and have no further specification should be dropped from the sample. This recommendation was to be extended to all the SITC sections. 19. SITC section 3: Mineral fuels, lubricants and related materials. It was recommended that there was need to check for outliers for very specific commodity items. 20. SiTC section 5: Chemicals and related products, n.e.s. It was recommended to exclude "other" commodities as in the case of commodities in the SITC section 2 and exclude outliers for specific items.

ECA/DISD/4.2d/2(v)/1999 Page 4 21. S1TC section 6: Manufactured goods classified chiefly by material. It was clear that there was nothing particularly new as regards to the commodities in the sub-section of manufactured goods, textiles and iron and metallic goods. The general recommendation made earlier would also apply to this section. 20. SITC section 7: Machinery and transport equipment. It was observed that size and quality of machinery can affect unit values. Some of the machines were meant for industrial use and others for household use. In addition, there are spare parts, second hand items, etc. which could also affect unit values. A careful examination of this section was called for 21. SITC section 8: Miscellaneous manufactured articles. It was recommended to consider a number of "other" categories, as long as their unit prices for the last three years were in the right direction. 22. Regarding examination of unit values, it was recommended to use unit price ratio comparisons for the three years 1993, 1994 and 1995 to lie between 0.8 and 1.2 (20% margin) and/or 0.7 and 1.3 (30% margin). (iii). Methodology for compiling trade indices 23. Concerning the methodology which was proposed by the CBS for the calculation of trade indices, the mission examined it in detail and compared it with the methodology recommended by the United Nations in this area. The two methods were found to be in harmony. 24. It was therefore recommended to retain methods already adopted for the computation of SITC section trade indices and the overall trade indices. In short a separate Laspeyres index for total imports and export and re-exports by section should be calculated and for the overall index, the geometric mean of the Laspeyres and Paasche indices should be calculated (Fisher Ideal Index). (iv). Splicing and linking/chaining 25. Full explanation with examples on splicing and linking/chaining of index numbers is reproduced in Annex II of this report. (v). Examination of specific problems: SITC section 0 26. Food items are only imported when there is short supply in the country. The opposite applies to exports. SITC section 1 27. For Brandy, Wines and Cigarettes, several brands are imported/exported.

ECA/D!SD/4.2d/(v)/1999 Page 5 SITC section 4 28. Unit of palm oil imported has problems - customs seem not to provide correct quantities of imports, thus affecting the unit price. SITC section 5 29. Lack of specification resulting into lack of comparability. SITC section 6 30. Textiles units are in kgs, sq meters, no's, etc. (vi) General problems 31. Further splitting of commodities resulting into lack of comparability since the two items i.e. at base period and current period are not the same. SITC section 8 mainly suffers from this problem. 32. Importation of low quality or substandard goods lowering the unit price. 33. Handling of subsidized goods, Initially, and after removal. 34. Under valuation of imports to avoid paying high import taxes. 35. Representation of heterogeneous commodities, (vii) Summary 36. The following is a summary of the discussion that took place: (a) The items in the first selection of commodities from which will be drawn the final items of commodities for inclusion in the index was reviewed; (b) The criteria for the first selection of commodities was as follows: Imports (a) Value not less than one million Kshillings; (b) Sources at least 5 countries. Exports (a) Value 20,000 Kshillings and above; (b) Sources at least 5 countries. (c) Problems and solutions: Variation in unit price in time-series was solved by removing outliers and considering average unit price for the remaining import or export items?. The CBS to recommend to the Statistical Office of Kenya Revenue Authority to further split of some items e.g. maize into ordinary maize for food and seed maize. Subsidies on goods would not pause a major problem since most countries including Kenya were moving towards market forces in control of the management of their economies. In

ECA/DlSD/4.2d/2(v)/1999 Page 6 price ratio comparisons for the three years 1993, 1994 and 1995 to lie between 0.8 and 1.2 (20% margin) and/or 0.7 and 1.3 (30% margin). Secondly when using this method to select items, the groups should be adequately represented. Other categories. For SITC sections 0 to 7, all "other" categories that did not have additional specifications other than "other" to be excluded. For SITC section 8, consider a number of these other categories as long as their unit prices for the last three years were in the right direction. (d) Methodology: The methodology in use was in harmony with the recommendations of the United Nations in this field. It was recommended to retain methods already adopted for the computation of SITC section trade indices and the overall trade indices. Recommendations: It was further recommended that the CBS should consider use of the chain method for linking the indices. If possible the 1994 index should be spliced with the previous 1982 based indices. A- Kenya - Part II (i) Introduction 37. A mission was undertaken to the Kenya Central Bureau of Statistics (CBS) from 1-5 December 1997 as a follow-up to an earlier mission which was undertaken to the CBS from 14-15 November 1996. 38. The objectives of the mission were: to examining further the selection of items; examining how representative the index was at the Section, Group, Division and item levels - sample coverage; calculate weights at the Section, Division and Item levels; to check whether re-exports data should be included in the export index; to revisit the decision of making 1994 as the base year of the trade indices; and to discuss the technical calculation of the index numbers. (ii) Selection of items 39. "other" items which were not specified were removed from the original listing of imports data and the same was being done in the case of exports data. There was a problem of lack of specification of items in group 7: Machinery and Transport Equipment. This was partly been due to the non-homogeneity of items in this group and as such may require special treatment in the assignment of weights. In all cases items unit prices outside acceptable range i.e. 30% margin, were examined further by removing outliers. (iii) Sample Coverage for Imports 40. The following 1994 sample was arrived at and its coverage at the Division, group and Item for each Section of the SITC is presented in the table below:

ECA/DISD/4.2d/(v)/1999 Page i Sample coverage for i mports Division Group Item Section 0 SITC Groups Sample 8 SITC Divisions 36 Sample 1b Original listinq 41 Sample 26 1 2 9 2 2 8 4 36 l6 12 15 47 10 30 3 9 2 11 4 22 14 4 5 4 3 3 9 4 33 4 24 13 295 8 91 6 9 / 52 28 406 80 7 9 8 50 2b 243 49 9 8 51 41 9 41. Section 7 was under further consideration particularly with respect to the inadequate representation of the groups. Groups of items such telephone sets, locomotive engines and airaaft spare parts which have a large contribution to total trade in terms of value, appeared not to be adequately represented. 42. The Groups and Divisions weights were also determined, (iv) Sample coverage for exports 43. The sample was not arrived at during the time of the mission, however a preliminary examination of the unit prices and whether they fall within or outside acceptable range showed the following results (see table below): Preliminary exports item coverage Items Section Original listing 0 68 1 9 2 33 3 15 4 5 5 43 6 103 r 16 8 66 * To be considered for further examination. Sample (likely) 48 5 16 9 3 17 42 0 11

ECA/DISD/4.2d/2(v)/1999 Page 8 44. The above results require further examination to ensure that the sample was representative enough particularly at the group level. Section 7 which had no sample was being further examined to see if this was due to lack of specification or outliers existing in the data. (v) Assignment of Weights 45. Assignment of weights for imports at the Section, Division and Group levels was completed. As for Exports, assignment of weights at the Section, Division and Group levels was yet to be done. (vi) Inclusion or exclusion of re-exports data in the export trade indices 46. To enable a thorough examination of this, the ratio of domestic exports to total exports by SITC Section was calculated for 1993, 1994 and 1995 trade data. The results of this calculation are shown in the table below: Ratio of domestic exports to total exports Year SITC Section 1993 1994 1995 0 1.00 0.99 0.98 1 0.95 0.96 0.95 2 1.00 0.99 0.99 3 1.00 0.99 0.99 4 0.99 0.99 0,95 5 0.96 0.94 0.94 6 0.99 0.98 0.97 7 0.48 0.45 0.47 8 0.97 0.97 0.92 46. It is clear from the above table that for all the SITC Section except SITC 7, it does not matter whether we consider total or domestic exports only in the index. 47. In the case of Section 7, it is clear that about 50 per cent of the total exports are re-exports. Therefore, if domestic exports only are considered in the index, this will represent about 50 per cent of total exports. On the other hand if total exports are considered, this will include about 50 per cent of re-exports. A lot depends on whether the situation regarding re-exports is likely to continue in the future. If indeed this is a short-term phenomena, then it is better to consider only domestic exports in the index. 48 An examination of the export items under Section 7 showed that they include the following: Tractors, mechanical shovels, food processing machinery, packing and wrapping machinery, machinery parts, loud speakers, passenger vehicles of cylinder capacity, parts and accessories or motor vehicles, etc.

56. Trade statistics of South Africa are based on the "Special Trade System". This means that imports which pass through a Customs bonded warehouse are included only at the time they are cleared for home consumption rather than at the time they enter the warehouse. ECA/DISD/4.2d/(v)/1999 Page 9 (vii) Revisit the decision of making 1994 as the base year of the trade indices 49. The choice of 1994 as base year was influenced by the fact that the external sector of Kenya was liberalised in 1993. 50. The choice of 1994 as the base year is becoming more acceptable particularly in view of the fact that there is nothing particularly unusual in terms of the behaviour of imports and exports statistics and the exchange rates normalised in 1994. (viii) Technical calculation of index numbers 51. As agreed upon earlier the technical calculations will follow the methodology described in the earlier report of the 1996 mission. B. South Africa Towards re-engineering the compilation of trade indices (i) Introduction 52. A technical mission was undertaken to Statistics South Africa from 23 to 27 November 1998. The objectives of the mission were to assess and evaluate the methodology adopted by South Africa for the compilation of international trade indices. 53. The major stakeholders for South African trade statistics are: The South African Revenue Services (SARS), Customs and Excise; Statistics South Africa (SSA); South African Reserve Bank (SARB); and Department of Trade and Industry, Government of South Africa. 54. The main source of trade statistics are the documents which exporters and importers file with Customs authorities prior to importing or exporting commodities. imports or exports of commodities are required to be reported irrespective of value. For statistical purposes the major items collected through customs documents are commodity, country of origin and destination, quantity, and value of exports or imports. Both omports and exports are valued on free on board (f.o.b) basis. Imports valued on cost, insurance and freight (c.i.f) basis are also available. To promote the comparability of international merchandise trade statistics, international recommendations of the United Nations state that imports be valued on c.i.f basis while exports be valued on f.o.b basis. 55. South Africa is a member of the South African Customs Union (SACU). The membership of SACU also include Botswana, Lesotho, Namibia and Swaziland. For customs purposes members of SACU are considered provinces of South Africa and as such no customs documents are filled for goods passing between these countries.

64. Accoring to information available at Statistics South Africa, the staff of the Rl nf Pavmpntc nivicinn at the.qarp alroarlw unrlor+aka on^lwele* «f (-.«tr«rj«^i«*«eca/disd/4.2d/2(v)/1999 Page 10 57. The commodity classification used by Customs and Excise for collecting trade data is the Harmonised System (HS), at eight digit level, six normal digits and two additional ones to take care of South Africa special needs. (ii) Processing and dissemination 58. The initial processing and dissemination of trade data is done by the SARS. The SARS uses its mainframe computer for processing trade data. By law, details of all imports and exports are to be submitted on paper to customs authorities. imports data are captured electronically at the ports while all exports data are sent to Pretoria for data capture. There is usually a time lag of one day for these transanctions between the ports and Pretoria. Imports and exports from other SACU members are also captured in Pretoria. SARS has plans to get all data captured at the ports and thereafter submitted to Pretoria. 59. Preliminary monthly trade figues are produced after 3-4 weeks from the reference month. After checking for errors the final figures are expected two month after the reference month but this sometimes takes 3 or 4 months. 60. Two monthly publications are relesed by the SARS. One publications presents the foreign trade statistics of South Africa while the other publication presents foreign trade statistics of the other SACU members -Botswana, Lesotho, Namibia and Swaziland. Final annual figures are produced with a time lag of about one year. (iii) Compilation of foreign trade indices at Statistics South Africa 61. Price and volume indices for exports and imports for periods up to and including January 1997 were produced and released by Statistics South Africa by the Chief Directorate of Economic Statistics and Surveys. The base year for these indices was 1988. The data which was used for compiling the indices originated from Customs and Excise, SARS with the exception of Gold whose data was supplied by the SARB. The release of these statistics was discontinued due to: the requirement of budgetary cuts within the office; the time lag in producing the publication was not acceptable; and there was a lack of staff experienced in producing these statistics. 62. In addition to many other users, the SARB used the data extensively to deflate Balance of Payments data to constant series. Because of delays in the release of the indices, the SARB estimated the indices until the actual indices were released by Statistics South Africa. Since the discontinuation of the release of the indices by Statistics South Africa, the SARB has continued to estimate price deflators. 63. Regarding the method used to calculate the indices, for Volume indices, the Laspeyres (fixed weights) formulae was used and for the Unit Value indices (price indices), the Paasche (moving weight) formulae was used. The Terms of Trade index was calculated as the ratio of Unit Value index of exports to the Unit Value index of imports expressed as a percentage. The calculation of the indices used software developed on Statistics South Africa mainframe computer.

ECA/DISD/4.2cl/(v)/1999 Page n which they receive from the SARS, any errors which they detect are cross checked with the SARS. These data are used to estimate trade indices on impots and exports for purpose of deflation. 65. Statistics South Africa was ready to provide to SARB computer programs which they used to calculate the trade indices, at the same time the methodology used to calculate the indices was due for revision. (iv) Compilation of the trade indices at Statistics South Africa 66. Three factors were said to have led to the present situation. Firstly the late receipt of data from Customs and Excise; turn over of staff; and lack of adequate knowledge of the methodology involved in the technical calculation of the index. 67. In describing the method used to cacluate the indices, it was stated that computer programs were already established on the mainframe computer of Statistics South Africa and all that was required was to input the data and let the computer undertake the calculations. Monthly data was supplied by SARS, preliminary figures to start with and later the final figures were made available. 68. The base year was indeed 1988 and the Unit Value and Volume indices were calculated. Seasonal adjustments were applied to the data when a whole years data was available. 69. The detailed methodology of the way the index was calculated including the weights which were established at the initial setting up of the index was not available. However what was reported was that the index was calculated monthly and also yearly and that the indices of Volume and Unit Value were seasonally adjusted and available for the Standard International Trade Classification (SITC) categories. This meant that the programs at SSA converted the data from HS to SITC for the purpose of calculating the indices and other analytical purposes. 70. It was suggested that if indeed the calculation of trade indices was to be underatken by Statistics South Africa, the whole methodology will have to be reengineered. This may also apply if another office was to undertake the calculation of the the trade indices. 71. The mission explained that software now exist for calculation of indices, in particular the Statistical Office of the European Communities (EUROSTAT) have through their software EUROTRACE for analysis of trade data recently come up with a software/module for calculating trade indices. (v) Other organisations involved in trade statistics South African Reseve Bank 72. The mission had an opportunity of discussing with the SARB and their work in the area of trade indices. The following was the understanding of the situation at their end:

ECA/DISD/4.2d/2(v)/1999 Page 12 Trade data is received monthly from Customs and Excise, SARS in electronic format; Trade data is provided by SARS in Harmonised System codes; Trade data is not converted into any other classification by the SARB, but when errors or inconsistencies are spotted, these are querried with the SARS; 73. The SARB calculates Unit Value (price) indices and Volume indices and also calculate indices for selected commodities for intenral use. Some of the data in particular the Balance of Payments get published in the SARB official publications; 74. The indices are used in the Balance of Payments for deriving constant price series as per the requirement of the International Monetary Fund (IMF); 75. The methodology for the calculation of the trade indices was not described in detail by the SARB, the mission however requested that it be sent to Addis Ababa.. South African Revenue Services, Customs and Excise 76. The South African Revenue Services, Customs and Excise collects trade data through the bills of entry completed by importes and exporters. SARS process the data and produce a preliminary report on trade statistics within 3-4 weeks after the end of the month in accordance to the 23 sections of the HS. After 2 months (this sometimes takes 3-4 months) when the data is edited they are in a position to produce the final monthly data which they supply to Statistics South Africa, SARB, Industry Development Cooperation (IDC), etc. These data are also printed in a Monthly Abstract of Statistics. At the end of the year the yearly data is put togerther ready for transmission to the United Nations, etc. 77. One problem which was explained was that Customs collect the bills of entry for revenue purposes. As a result because there is no customs duty for exports they tend not to pay too much attention to the details on exports. They however tend to pay more attention to details of imports because revenue is involved. 78. On software which they use, it was explained that they do not use Asysuda software, instead they use their own system which they found suitable. The data is also processed on their maiframe computer. 79. On the delay in submitting the data to Statistics South Africa, they explained that they try their best and believe that with the staff which they have (10 in all) they are able to provide the data in a timely manner. 80. From January 1999, it is expected that inta-sacu trade will be be collected and be made available, since for all goods leaving or entering the SACU, there will exist a customs declaration form.

ECA/DISD/4.2d/(v)/1999 Page 13 (vi) Recommendations 81. As a result of the technical study, the following recommendations are made: The stakeholders of South African trade statistics should come together and clearly outline responsibilties, and either reafirm or reassign them as necessary; Statistics South Africa should take the lead in establishing a Task Force on Trade Statistics, the membership of which should include: SSA, SARS, SARB, and Ministry of Trade Industry. This task force should meet periodically to review among other things: methodology for data collection, analysis and dissemination and coordination issues of South Africa trade statistics; Due to unavailability of detailed methodology on the compilation of South African trade indices for the period 1988 to January 1997, which was said to have been caused either by lack of or loss of documentation, it is recommended that work on the compilation of trade indices should be re-engineered; The responsibility for the work on trade indices should be discussed by the Task Force on Trade Statistics which has been recommended in bullet number two of this paragraph; The United Nations should assist in making available methodologies which have been adopted by other African countries in the area of trade indices. C. Mauritius Fully developed set of indices (i) Introduction 82. The mission to Mauritius Central Statistics Office was undertaken from 16 to 20 November 1998 with the objective of assessing and evaluating the methodology adopted by Mauritius for the compilation of foreign trade indices. The mission findings were that Maritius compiled and published a number of quarterly indices as follows: Unit value index (imports), base year 1992; Unit value index (domestic exports), base year 1992; Export processing Zone (EPZ) index (Unit value index for exports of EPZ), base year 1992; Export price index (EPI), base year 1993. 83. The mission was provided with excellent methodological documents on the above indices and are produced below. (ii) External trade statistics - general aspects

ECA/DISD/4.2d/2(v)/1999 Page u Sources of data 84. The basic trade data on which the unit value indices are based are compiled by the Central Statistical Office on a monthly basis from the customs returns filed by exporters and importers. These data are supplemented with direct information from the following organisations: The State Trading Corporation for government import of petroleum products and rice The Mauritius Sugar Syndicate on exports and imports of sugar The Mauritius Marine Authority for import and export of marine vessels Air Mauritius for import and export of aircraft and parts Methodology of compilation 85. Trade figures are compiled according to the General Trade System as recommended by the United Nations. Using the national boundary as the statistical frontier, the General Trade System is a record of all goods entering (imports) and leaving the country (exports). However, for the sake of comparability and continuity, transactions of the Freeport are not presently included in the trade statistics. 86. Imports include goods brought in directly for home consumption and goods imported into customs bonded warehouses. 87. Exports cover exports and re-exports. Domestic exports are defined as goods of national origin. Re-exports are goods which are exported in the same condition as imported or after undergoing minor operations which leave them essentially unchanged. 88. The following are excluded in the compilation of trade statistics: IMPORTS Imports of a value not exceeding Rs. 1,000 Transhipment Re-warehousing Returned goods, postage stamps, bank notes and coins Empty containers Personal effects Goods imported by foreign embassies and duty free shops Fish and other sea products landed in Mauritius from the High Seas by Mauritian vessels

ECA/DISD/4.2d/(v)/1999 Page 15 Goods returning after repairs abroad.. EXPORTS exports of a value not exceeding Rs. 1,000 Returned goods, postage stamps, bank notes and coins Empty containers Goods sent abroad for repairs Personal effects Goods exported by duty free shops. Commodity Classification 89. Commodities are coded according to the Harmonised Commodity Description and Coding System Nomenclature (HS). However, for the purposes of economic analysis and to facilitate international comparison of trade by commodity data, the commodities are also classified according to the Standard International Trade Classification, Revision 3 (SITC Rev. 3). In order to cater for national needs, both the HS and the SITC Rev.3 have been extended to eight and seven digits respectively. Processing 90. The CSO relies on the services of the Central Information Systems Division (C.I.S.D) of the Ministry of Telecommunications and Information Technology. Transactions for a particular month are received on diskettes from the Customs Department and then forwarded to the Central Information System Division (C.I.S.D.) for computer processing. The latter then supplies the CSO with computer listings which group all similar items falling under a given classification (HS) with all relevant fields. In addition it gives the unit value of each transaction which is the basic check of data quality. As the listing shows individual entries classified by each single tariff item number, unit prices can be compared with each other. Whenever the unit price is found to be "out of range", the officers accede to the bill of entry for verification and necessary amendments are sent to the C.I.S.D. for the cleaning of the data. Subsequently statistical tables are produced on a quarterly and yearly basis. (iii) Foreign Trade Unit Value Indices Definition and structure 91. The unit value index is an estimate of the unit values of imports/exports in any given period relative to the unit values of those imports/exports in a predetermined (base) year.

ECA/DISD/4.2d/2(v)/1999 Page 16 92. A major problem of such an index is that a change in the average unit value index does not necessarily signify that the fluctuation was due to a difference in price, but could well be attributable either to a shift in the quality of a product or to a change in the product mix within the commodity code or still a change in the country of destination/origin, classified nevertheless under the same item code. 93. The unit price indices, like all indices using fixed weights, are affected by changing structure of the economy. Thus after a certain time, they become less representative of the actual price movements they are supposed to monitor. In order to keep the indices more up-to-date so as to reflect changes in trade structure, it is necessary to revise them periodically, say every five years. 94. The CSO presently compiles on a quarterly basis three series of unit value trade indices, viz Domestic exports unit price index Imports unit price index Export Processing Zone (EPZ) exports unit price index. 95. Unit value indices have been calculated as far back as 1964. The present ones are being compiled using the year 1992 as base. Since then, there has been a shift in the importance of goods imported or exported. It is therefore necessary to update the basket to represent the actual commodities imported or exported. Hence the present revision which uses 1997 as base year. 96. For the first time, countries of origin/destination are taken into consideration in the calculation of the revised indices, thus improving the present method of compilation. A country is included if the value of transactions for that particular item during the base year, that is 1997, represents at least 1% of the total value of imports/exports of that commodity and the commodity has been traded for at least two quarters during the base year. Uses of foreign trade indices 97. The following are the uses of forein trade indices; To analyse the effects of devaluation To provide deflators for national accounts To make budget revenue projections and macro-economic forecasts They offer the possibility of breaking down the global value of imports and exports into their price component and subsequently volume component. They are also of importance in the evaluation of the terms of trade They provide quarterly measures of the trend in the price of the country's exports and imports.

-Sr='^i*K^ ECA/DISD/4.2d/(v)/1999 Page n (iv) Calculation (a) Commodity level 98. For each country the product P1Q0 and PoQo is calculated and summed over all countries represented to give the values for the commodity. (b) Group level 99. An index is calculated for each group using the Laspeyre fixed weight formula the sum being extended over all commodities represented in group i. (c) Division level 100. The index for each division is derived as a weighted average of the indices of al! groups within that division. The formula used is mmmmmmam Where log) = Index for division j Wj = Weight of group i Igo) = Index of group i (d) Section level 101. The index for each section is calculated as a weighted average of the indices of all divisions within that section using the formula Where ls«= Index for section k Wj = Weight of division j Idj = Index of division j within section k (e) Overall index

ECA/DISD/4.2d/2(v)/1999 Page is 102. Finally the overall unit value index is derived as a weighted average of the indices of the various sections. The formula is Where I = Overall unit value index Wk = Weight of section k lsk = Index of section k (v) DOMESTIC EXPORTS Introduction 103. The Export Unit Value Index shows price trends for exported commodities. The price used in the compilation of the index is the FOB (Free on Board) price of the commodity exported. This index is used as a deflator for exported goods at current prices and also in the calculation of the Terms of Trade index. 104. However, as mentioned earlier, this unit value index does not reflect pure price changes since it is affected by factors like product mix, changes in quality and country of destination, seasonality and discontinuity. For this reason, the CSO also compiles a quarterly Export Price Index (EPI) based on prices of well specified products determined through direct surveys of exporters. Coverage 105. This index covers 91 % of total exports. The following were excluded: items with an annual customs value below Rs. 10 million Sections 1, 3, and 4 of the S.I.T.C (rev. 3) which include commodities which are rarely exported. Section 9, which by its very nature of grouping commodities and transactions not classified elsewhere and none of which had an annual value of Rs. 10 million or more has also been excluded, as recommended by international organisations. These four sections together represent less than 0.1% of the total exports. Works of art, wood maquetry, jewellery and precious stones for which quantities are not given, thus making it impossible to calculate unit price. 106. Table A shows the coverage by S.I.T.C sections for 1992 and 1997 Major changes have occurred only in section 2 {crude materials, inedible, except fuels) and section 7 (machinery and transport equipment). In the case of section 2t the coverage

ECA/DISD/4.2d/(v)/1999 Page 19 Division 26 (textile fibres, other than wool tops and other combed wool, and their wastes) and Division 27 (crude fertilisers other than those of Division 56, and crude minerals). Total exports of goods within these two divisions amounted to Rs. 103 million in 1997, representing 35% of total exports of Section 2. Commodities of Division 26 have been excluded because they all had an export value below Rs. 10 million whereas in the case of Division 27 the goods have not been included because quantities are not available. 107. On the other hand, the coverage of Section 7 has increased from 42% in 1992 to 67% as a result of the increase in the share of exports of industrial machinery and equipment. 108. A detailed table on the coverage by section, division and group will be included in the final document to be released by the CSO, Mauritius. Table A - Coverage by S.i.T.C sections (1992 and 1997) 0 - Food and live animals 9,162 9,008 99 1 - Beverages and tobacco 17 2 - Crude materials, inedible, except fuels 294 164 56 78 3 - Mineral fuels, lubricants and related materials 4 - Animal and vegatable oils, fats and waxes 11 5 - Chemical and related products, n.e.s 149 104 70 69 6 - Manufactured goods classified chiefly by materials 2,170 1,459 67 72 7 - Machinery and transport equipment 105 70 67 42 8 - Miscellaneous manufactured articles 19,983 18,159 91 87 TOTAL 31,891 28,964 91 90 Weighting Scheme 109. The weights at the various levels of aggregation was made proportional to the 1997 export values of the section, division and groups as the case may be. Table B below shows the weight attributed to each section covered whereas a detailed distribution of weights by section, division and group will be included in the final document to be released by the CSO, Mauritius.

ECA/DISD/4.2d/2(v)/1999 Page 20 Table B - ocation of weights by section (1997) 0 - Food and live animals 2875 2 - Crude materials, inedible, except fuels 92 5 - Chemical and related products, n.e.s 47 6 - Manufactured goods classified chiefly by materials 681 7 - Machinery and transport equipment 33 8 - Miscellaneous manufactured articles 6272 TOTAL 10000 (v) EXPORT PROCESSING ZONE Introduction 110. The Mauritius Export Processing Zone (EPZ) plays an important role in the economic development of the country. The striking performance of the Mauritian economy is based on the continued and sustained encouragement provided to exportoriented industry which has now assumed the role of leader on the economic front and relegated sugar into the background. For instance, exports of EPZ represented 72% of totai domestic exports in 1997. It is for this reason that a separate unit value index for exports of the EPZ is compiled. Coverage 111. The index covers 86% of the total exports of the EPZ. Most of the EPZ exports (85%) fall under section 8 of the S.i.T.C (rev. 3) "Miscellaneous manufactured articles" which has been covered to an extent of 89% in the index. 112. The following were excluded: items with an annual customs value below Rs. 10 million Sections 1 to 5 and 9 of the S.I.T.C (rev. 3) which include commodities which are rarely exported or those falling under "commodities and transactions not classified elsewhere" and none of which had an annual value of Rs. 10 million or more. These six sections together represent less than 1 % of the total exports Works of art, wood maquetry, jewellery, precious stones and other commodities for which quantities are not given, thus making it impossible to calculate unit value.

ECA/DISD/4.2d/(v)/1999 Page 21 Table C below shows the coverage of the various S.I.T.C sections compared to 1992. Table C - Coverage by S.I.T.C sections (1992 and 1997) mmmmssmmm 0 - Food and live animals 1,249 967 87 1 - Beverages and tobacco 0 2 - Crude materials, inedible, except fuels 90 57 3 - Mineral fuels, lubricants and related materials 4 - Animal and vegatable oils, fats and waxes 5 - Chemical and related products, n.e.s 35 91 6 - Manufactured goods classified chiefly by materials 2,043 1,335 65 53 7 - Machinery and transport equipment 96 57 59 53 8 - Miscellaneous manufactured articles 19,536 17,379 89 96 TOTAL 23,050 19,738 86 93 Weighting Scheme 113. The weights at the various levels of aggregation were made proportional to the 1997 export values of the section, division and groups as the case may be. Table D below shows the weight attributed to each section covered whereas a detailed distribution of weights by section, division and group will be included in the final document to be released by the CSO, Mauritius. Table D - ocation of weights by section (1997) Section 0 - Food and live animals 6 - Manufactured goods classified chiefly by materials 7 - Machinery and transport equipment 8 - Miscellaneous manufactured articles TOTAL Weigwll 545 891 42 8522 10000

ECA/DISD/4.2d/2(v)/1999 Page 22 IMPORTS Introduction 114. The considerable economic and structural changes in the Mauritian economy and the liberalisation of trade have led to a greater diversity and an increase in the number of products imported. For instance, in 1997 about 5000 different S.I.T.C item groups have been imported. Furthermore, many items imported are seasonal and subject to frequent changes in specifications. Coverage 115. In view of the problems mentioned above, it is not possible to maintain a coverage as high as that of exports. The groups are heterogeneous and countries from which a particular product is imported may change from one month to another. This leads to a wide dispersion in the unit values for a particular item group. 116. The index covers 68% of total imports and excludes the following: items with an annual c.i.f. value below Rs. 10 million Works of art, wood maquetry, jewellery and precious stones and other products for which quantities are not given, thus making it impossible to calculate unit price. Aircraft, marine vessels and parts 117. Table E below shows the coverage of the various S.I.T.C sections compared to 1992. The low coverage of Section 7 "Machinery and Transport Equipment is explained by the exclusion of aircraft, marine vessels and industrial machines imported on an irregular basis. 118. In section 8 "Miscellaneous manufactured articles", Division 87 which includes professional, scientific and controlling instruments, not elsewhere specified, has been under-covered (30%) because of the exclusion of certain expensive instruments which are not imported regularly. Besides, this section contains many item groups each with a total c.i.f. value of less than Rs. 10 million.

ECA/DISD/4.2cl/(v)/1999 Page 23 Table E- Coverage by S.LT.C sections (1992 and 1997) wmmmmmsmmmm 0 - Food and live animals 6,091 5,103 70 1 - Beverages and tobacco 262 207 79 75 2 - Crude materials, inedible, except fuels 3 - Mineral fuels, lubricants and related materials 4 - Animal and vegatabie oils, fats and waxes 5 - Chemical and related products, 1,567 1,071 68 67 3,471 3,430 99 98 484 436 90 81 3,340 2,302 69 73 n.e.s 6 - Manufactured goods classified 14,947 11,252 75 56 chiefly by materials 7 - Machinery and transport equipment 11,702 5,292 45 43 8 - Miscellaneous manufactured 3,804 2,254 59 54 articles TOTAL 45,668 31,347 68 60 Weighting scheme 119. The weights at the various levels are made proportional to the c.i.f. values of the section, division and groups. Table F below shows the allocation of weights by section whereas a detailed distribution by section, division and group will be included in the final document to be released by the CSO, Mauritius. Table F - ocation of weights by section (1997) at 0 - Food and live animals 1,334 1 - Beverages and tobacco 57 2 - Crude materials, inedible, except fuels 343 3 - Mineral fuels, lubricants and related materials 760 4 - Animal and vegatabie oils, fats and waxes 106 5 - Chemical and related products, n.e.s 731 6 - Manufactured goods classified chiefly by materials 3,273 7 - Machinery and transport equipment Z563 8 - Miscellaneous manufactured articles TOTAL 833 10,000

ECA/DISD/4.2d/2(v)/1999 Page 24 (vii) Export price index Methodology for the construction of the Export price Index (EPI) Introduction 120. The export price index is a measure of the price change of domestically produces Mauritian products shipped to other countries i.e. it excludes re-exports. In addition to the overall index, separate sub-indices are also calculated for different product categories at more deytailed level. Comparison with the unit value index 121. The unit value export index provides only proxy measure for the price changes of exported goods. It is derived from value and quantity data of products or groups of products obtained from trade returns and is thus highly dependent on the degree of homegeneity of the groups of products considered. On the other hand, the EPI measures pure price changes and is based on actual price measurements of relatively more homogeneaou group of products. It requires elaborate specifications of products and takes account of the main price determining factors such as country of destination, quality, mode of transport and mode of payment. Scope and classification 122. The EPI covers all domestic exports, the most important commodities being sugar, molasses, tea and textile fabrics and articles of apparel and clothing accessories produced by the EPZ sector. 123. The index is based on the nomenclature of the Standard International Trade Classification of the United nations (SITC Rev.3). Separate sub-indices are produced for each SITC section and for more detailed groups where possible. Selection of items and firms 124. A representative sample of 42 firms has been selected from trade declarations submitted to Customs and Excise Department in 1993. The sample was drawn from exporting firms on the basis of regularity of their trade and their volume of exports. A total of 119 items (Entry Level Items, ELI's) which represent 32 product groups are priced from the selected enterprises. Weighting of items and firms 125. The weights are derived from the 1993 domestic exports. Weights assigned to each section and group are based on their export values in 19993. Each product or group of products selected for pricing purposes represent all products that fall within that weight group.

ECA/DISD/4.2d/(v)/1999 Page 25 Base and reference period 126. The base price as well as the reference period for the EPI is calendar year 1993. Bench mark data for the year 1993, based on the trade returns, have been used to derive weights. The selected firms have suplied prices of selected commodities for each of the 12 months in 1993. The yearly average price of each product has then been worked out and is used as the base price for that product. Price collection 127. Prices are directly collected from co-operating firms. As far as possible prices quotations are reported on a free on board (f.o.b.) basis and are mostly contract prices. Each reporting firm is visited once every quarter but prices are supplied on a monthly basis and are averaged for each quarter. Sugar and molasses 128. Due to their specificity, sugar and molasses have been treated differently. Sugar and molasses produced during a crop year, which normally extends between July and June of the following year, are usually exported during that same crop year. Export prices of these commodities, sugar in particular, are mostly negotiated prices. These prices can be quite volatile between quarters depending on the destination of the shipment. Therefore, in order to eliminate these price distortions between quarters, the same average yearly price for the crop year is used for the four quarters comprising the crop year. The base price of these commodities for the year 1993 is the average price for crops years 92/93 and 93/94. Calculation of the EPI 129. As modified Laspeyres formula based on the weighted average of price relatives is used to calculate the EPI: the mathematical form of the formula is shown below: f -1J Where lot is the index for the period t compared to base period o Wj is the weight of the im element Poi is the base price of the ith element Pit is the base price of the ith element in period t (Pit/Pot) is the price relative of the ith element in period t relative to base period o (summation over all)

ECA/DISD/4.2d/2(v)/1999 Page 26 Uses 130. The primary use of the EPI is to deflate export trade statistics. It provides quarterly measures of price trends of Mauritian products sold abroad and can be used for calculating changes in the volume of exports. 131. It can also serve as a basis to assess the competitiveness of the Mauritian products in relation to price trends of common products of other countries with which Mauritius competes for markets. IV. Conclusions and recommendations 132. The case study has revealed a number of technical aspect that require to be considered by countries in Africa. 133. In the case of Kenya, it is clear that re-basing trade indices requires a careful study of the structure of trade and the commodities involved. Removing outliers to come up with a good representative sample is one of the considerations. Also the choice of the base period has to be decided upon on a basis of sound assumptions. The methodology to be adopted should to the extent possible be based on that which has worked elsewhere i.e. the United Nations recommended methodology. 134. In the case of South Africa, since work on the indices was halted due to factors beyond control of Statistics South Africa, there is need to continue the work based on the recommendations which were made by the Economic Commission for Africa technical mission. 135. In the case of Mauritius, a lot of work was already accomplished. The case of Mauritius could be used as a model for other countries to learn from. The achievement of Mauritius in producing many quarterly trade indices is to be commended. 136. In terms of recommendation arising from the technical study, the following can be stated: (a) Countries should to the extent possible attempt to compile trade indices; (b) This is perhaps and area through which technical co-operation among developing countries could be promoted, to ensure that techniques which have worked in one country are replicated in countries with similar condition; (c) It is recommended that study tours or attachment of officers to countries that have succeeded in compiling trade indices be encouraged. (d) Dialogue with users of data is encouraged to ensure that indices produces will be put to use before this extensive work is embarked upon. (e) Technical agencies should do their best to assist African countries in doing this work.

Annex>I ARTICLE: THE NEW TRADE INDICES 1977 VoK XV - No. 2 Price: Sh. 6/00 or 24/- per year-post free Central Bureau of Statistics Ministry of Finance and Planning

Thus, the Laspeyre index used is simply the sum of the price relatives of the selected commodities included in the index multiplied by appropriate weights. THE NEW TRADE INDICES Introduction 1. New indices have been constructed because imports and exports are now reported gross. Previously imports were reported net (i.e. imports from overseas that were transferred to Uganda and Tanzania were deducted). Gross imports include not only total imports from overseas but those from Uganda and Tanzania as well. Analagously, total exports include exports to overseas countries, transfers to Uganda and Tanzania, and re-exports (including imports from overseas subsequently re-sold to Uganda and Tanzania). 2. A second reason for constructing new indices was that improvements were needed in method. In particular, a method was needed that took into account the changing structure of trade; and that realistically measured price and quantum changes. The Need to Measure Price Changes 3. One object of these indices is to measure changes in the terms of trade. Since the terms of trade refer to price changes, and not to changes in average unit values, the indices are designed to measure price changes. 4. This need for the indices to measure price changes can be illustrated with a simple example. Suppose in a given year that the prices of all imports and exports have remained constant except those of cars; and that the price of every make of car has risen. Clearly the terms of trade have moted against Kenya. Suppose, however, Kenya imports a larger proportion of smaller cars (Minis), and lesser proportion of bigger cars (Mercedes). The average unit value of cars will fall, despite the price of each make having risen. If one was wrongly using the average unit values of all cars as an indicator of price changes, one would get the false result that the terms of trade had moved in Kenya's favour. But this is not so. What has happened is that there has been a change in consumer preferences. And this has led to a higher proportion of Minis being imported. Numerous other examples can be drawn from both the export and imports list. 5. If price indices arc to be reliable, it is essential that the unit prices of the items included in the indices should be as specific as possible. Changes in unit values will then reflect changes -in prices rather than changes in the proportions of various types of goods included in the different items. It is therefore the constant endeavour to refine these unit values, so that fictitious price changes resulting from a change in the import (export) mix of an item do not get incorporated in the indices. Methodology 6. Foreign trade is classified according to S.I.T.C. into the under-mentioned eight major sections: Section 0 Food and Live Animals. 1 Beverages and Tobacco. 2 Crude Materials. 3 Mineral Fuel and Lubricants. 4 Animal and Vegetable Oils and Fats. 5 Chemicals. 6 Manufactured Goods classified by Material. 7 Machinery and Transport Equipment. 8 Miscellaneous Manufactured Articles. Section Indices 7. A separate Laspeyrc price index is calculated for each section's total imports and exports and re-exports. The formular for such an index is: Index PliJ J «= woij... PoiJ where:, P] i J -- current price of commodity i in section J PoiJ --- base-year price of commodity i in section J woij Poi.T- QoiJ S PoiJ-OoiJ QoiJ -- - Quantity of commodity i in section.1 in base year.

Within Section Weights 8. The first step in calculation of within-section weights was to divide the component items into groups of dose substitutes, or groups of items containing similar raw materials. The idea of course was that the prices of items in such a group, since close substitutes, tend to move together. This was equivalent to stratification in a sampling survey. The importance of stratification in sampling is well known. It increases the efficiency of an estimator by reducing the within stratum variation. Similarly, stratification improves the quality of a trade index. It reduces the unit price variation. The collective weight given to such a group was the proportion in 1972, of the value of all imports (exports) in the group to the total value of all imports (exports) of the section. 9. The next step was to select an item, or items to represent the group. The general principle adopted was that the larger the relative value of the group, and the less homogeneous the component items, the larger the number of items taken to represent the group. Suppose commodities A, B and G have been selected to represent a group with a collective weight of 0.30; suppose, further, that the proportions imported or exported of A, B, and C are 0.22, 0.50 and 0.28 respectively (adding to 1.00). Then the weights allocated to A, B, and C are these proportions multiplied toy the group weight (0.30) i.e..066;.350 and.084 respectively. The items included in the indices and their within section weights are given in Table 3. For any section the sum of the individual weights equals unity. Some Problems 10. Every effort was made in selecting representative commodities for the reasons stated in paragraph 5 to make them as specific as possible. A comparison was made for each selected item of the variation in unit values of the imports from different countries. The spread in these unit values was frequently large. When the range exceeded 20 per cent of the average unit value of imports from all countries, the item was restricted to imports from a specific country. This was not done for exports. It was felt that the price for most export commodities was not likely to vary significantly with destination. 11. The difficulties of selecting representative commodities for the indices of sections 7 and 8 were particularly acute. Homogeneous items do not exist. For example, in section 7 the items selected to represent the first group include agricultural machinery, outboard motors, tractors machine tools, etc. None of these refers to a specific good. The unit value of outboard motors, say, could vary If larger more expensive engines were imported (even though the price of each type had remained unchanged). It seemed safer, in view of tine danger of fictitious price changes, not to give the items representing the group individual weights. Instead a geometric mean of the price relatives of the selected items in the group is first calculated; then the group weight is applied to this average price relative. With this method, it is hoped that the fictitious price changes will offset each other, thus allowing the true price change to emerge. This offsetting tendency might not operate if the selected items in the group were given individual weights. For a fictitious change in the price relative of an Item with a heavy weight could in this case badly distort the price index of the section. 12. It is even more difficult still to construct a price index for section 8. For only a few groups such as that for clothing is it possible to derive unit values for representative items. Quantities imported are not stated. Nor is there any season why the prices of goods in the remaining groups which, consist of a vast array of different types of manufactured goods should move in line with those of clothing. It seems likely -that their prices, in view of the variety of goods, will vary with those of manufactured goods generally. And this was the assumption that was made. Accordingly, the price index for manu facturing serves as a price relative and is multiplied by a weight reflecting the relative value of the groups for which representative items could be selected. Overall Price Indices 13. The price indices for the 8 different sections having been calculated, the next task is to combine them into one index. They are combined by multiplying them by section weights. But this is done in two ways. In the first case, the section weights are the proportions of the value of section imports (exports) to the total imports (exports) in 1972. In the second case, the section weights are again these proportions but derived from current year values. The two separate indices" for all imports (exports) are next combined into one index by taking their geometric mean. 14. To bo more specific, the two formulae used are: Base-year weights Index =2 Index 3. WoJ Where: Index J Price index for section J calculated as stated in paragraph 7. WoJ = Proportion of value of imports (exports) of section J to total value of imports (exports) in base year.