5. "SPECIAL PRODUCTS": APPLICATION OF THE METHODOLOGY - CASE STUDY RESULTS

This section presents results in two frames. First, analysis in the context of the criteria as evaluated through product, indicators and their trade policy regimes. Secondly, an evaluation of the indicators themselves, using factor analysis to test for interrelationships and possible reduction of the indicators into principal factors.

5.1 Indicator analysis results

The indicator analysis is based on secondary data available in the public domain for four countries - Belize, Egypt, Nigeria and Thailand. The national, agricultural and trade policy goals as presented in the WTO Trade Policy Review, National Development Plans and Agricultural Sector Plans and Policy Statements provided the background and context to the analysis.

Table 1 provides a summary of the nine indicators across which data were collected for each country. Four, three and two indicators were used for food security, livelihood security and rural development, respectively. The choice was driven by data availability.

Each agricultural product for the country in the FAOSTAT database was evaluated against the nine indicators above. Table 2 shows the number of products for each country evaluated in the context of the indicators and the number of products that qualified in the top thirty in at least one of the indicator categories. It also shows the number of products in terms of the number of indicators under which they qualified.

The total products produced (and exported) and imported for each country indicates the total number of products evaluated. Obviously, some indicators evaluated only imports or production (exports) and depending on data availability the numbers of products evaluated under each indicator varied. For each base variable/indicator the top thirty products generally accounted for more than 75 percent of the total activity reflected by the indicator. For example, the top thirty products under total calorie consumption or land harvested accounted for greater than 75 percent of calories consumed or land harvested. The top thirty products for each variable was considered the base set of possible products, and products that made that cut and identified with at least three indicators were chosen as the list for further analysis. Table 3 shows the products and the indictors for each country that were in the top thirty on at least three of the indicators. Annex 2 presents the country tables with the indicator values for the products qualifying with three indicators.

TABLE 1
Criteria and indicators against which products evaluated

¹ For the period 1985 - 2002.

TABLE 2
Number of products evaluated by country and number of products meeting 1 to >=3 indicators

TABLE 3
Products and their respective at least three related indicators

The FAOSTAT product descriptions were then mapped with the equivalent HS codes for the products qualifying against three indicators. This allowed evaluation of the trade characteristics of the product. Tables 4 and 5 show these results for Belize and Thailand given that only for these countries the information on the trade policy affecting the products were relatively more complete and available.

The trade policy information from Belize is particularly reinforcing of the indicator analysis approach adopted as 75 percent of the products (fifteen out of twenty) that arise from the indicator analysis are products that are exceptions (above) to Belize’s ceiling binding of 100 percent, products with a bound tariff of 105 or 110 percent. Further, all but one of the products has a tariff overhang of greater than 60 percent, the difference between Belize’s Common External Tariff and its ceiling binding. The twenty products translate into thirty-seven tariff lines, sixteen of which appear on the Schedule C (thirty-eight tariff lines) of exceptions. Most of the products that do not appear on the selected indicator list of three or greater appear on the list of products qualifying with one or two indicators for Belize. Thus, in the case of Belize a largely overlapping set of products are possible for Special Product identification. In the categories of the framework agreement perhaps as follows: food security (rice, maize, poultry, dried beans), livelihood security (sugar, citrus, bananas), rural development (pig meat, fruits, potatoes).

TABLE 4
Belize High Indicator Count Products and Trade Policy dimensions

Note: For products with multiple HS Numbers, tariffs presented in the table are average tariffs of those HS Numbers

Sources: WITS, World Bank; WTO

TABLE 5
Thailand High Indicator Count Products and Trade Policy dimensions

Note:

i) For products with multiple HS Numbers, tariffs presented in the table are average tariffs of those HS Numbers

ii) Since data on bound and applied tariffs are for different years, for some products applied tariffs may be higher than bound tariffs.

Sources: WITS, World Bank; WTO

TABLE 6
Egypt High Indicator Count Products and Trade Policy dimensions

Note: For products with multiple HS Numbers, tariffs presented in the table are average tariffs of those HS Numbers

In the case of Thailand eight of the sixteen products that arise out of the indicator analysis have either been designated for use of the SSG or have a TRQ or both (Table 5). The food security crops (rice, sugar, cassava, maize and chicken meat) are on the indicator list, with maize, chicken meat and rice all also having an SSG designation. These are also the crops that are prominent among the indicators under livelihood security and rural development and reinforce the expectation of overlapping product influence across the indicators and criteria. Some products that were also important under the two other criteria were coconuts and pineapples (livelihood security) and milk, cotton and groundnuts (rural development).

Table 6 and 7 show the high indicator count products and trade policy information for Nigeria and Egypt. For these two countries the trade policy information included is limited to tariff levels for the products.

In the case of Egypt, maize and rice are critical food security crops and show up among the high indicator count. The low bound and applied rates for maize is consistent with the fact more than 40 percent of total domestic consumption of maize, like wheat, is met from imports.

In terms of livelihood security, the indicator analysis points to sugar, potatoes, oranges, grapes and sesame seed in addition to cereals and fruits and vegetables as accounting for the majority of the area harvested. The link with employment for these crops is implicit.

For Nigeria the main products affecting food security are in the list of high indicator count products, particularly millet, cassava, maize, yams, sorghum, and cowpea. All of these products have the highest applied (100 percent) and bound (150 percent) rates, except maize which has a high applied rate of 70 percent. These products are among the major products contributing to calorie availability. In terms of livelihood security, the main products identified by the indicators are cotton, cocoa beans and groundnuts. These crops are particularly important for employment in rural areas as indicated by the share of land harvested. For Nigeria there is a considerable overlap between the criteria hence most of the food security and livelihood security crops contribute to rural development.

TABLE 7
Nigeria High Indicator Count Products and Trade Policy dimensions

Note: For products with multiple HS Numbers, tariffs presented in the table are average tariffs of those HS Numbers

TABLE 8
Low estimate of Tariff Lines for SP Exemption

An important result in the context of the negotiations for "Special Products" is also the number of tariff lines that might be potentially claimed for exemption. If all of the products classified with more than three indicator categories were evaluated by their equivalent tariff lines it is seen in Table 8 that based on the assumptions underlying the analysis here this could be as much as six percent of tariff lines. This is considered a low estimate as the analysis here is based mainly on national level aggregated data. When the product diversity across poorer regions of countries is considered it is expected that this number could be more than doubled.

5.2 Factor Analysis Results

Two of the four case studies were analysed using the factor analysis approach.^[11] Although the factor analysis and not the correlation matrix is the aim, it is nevertheless useful to firstly consider the correlation matrices reflecting the relationships between pairs of variables. Table 9 presents the correlation matrix for Egypt and Belize.

TABLE 9
Correlation Matrix for Indicators for Egypt and Belize *

* Elements in the principal diagonal are the squared multiple correlation coefficents (smc).

The coefficients of correlation express the degree of linear relationship between the row and column variables of the matrix (indicators of food security, livelihood security, and rural development). The closer to zero the coefficient, the less the relationship; the closer to one, the greater the relationship. A negative sign indicates that the variables are inversely related.

To interpret the coefficient, it is squared and multiplied by 100. This will give the percent variation in common for the data on the two indicators that the coefficient represents. Thus, in Table 9, for Egypt the correlation of.509 between growth rate in production and production variability means that 26 percent (.509² X 100) of the variation in the total number of commodity values used to evaluate these two indicators are in common. In other words, if one knows the commodity values used to construct one of the two indicators one can produce (predict, account for, generate, or explain) 26 percent of the values on the other indicator.

Consider the correlation of.873 between growth rate in production and growth rate in imports as another example. Within the given framework this correlation implies that 70 percent (.873² X 100) of the growth in imports of the commodities can be predicted from their production growth rates. Thus, assuming that the sample of commodities is random, if an additional commodity were randomly added to the sample and only its growth rate in production were known, then its variability could be predicted with 26 percent certainty and its import growth within 70 percent of its true value. For Belize (Table 9b), the correlation between import dependency and self-sufficiency is.95 indicating that 90 percent of the degree of self-sufficiency can be predicted from information on import dependency.

TABLE 10
Factor Matrices - Egypt and Belize

The principal diagonal of the correlation matrix is indicated in italics. The principal diagonal usually contains the correlation of a variable within itself, which is always 1. Often, however, when the correlation matrix is to be factored (using the common factor analysis model), the principal diagonal will contain communality estimates instead. In this context the communality estimates are across indicators/variables. In the reference to communality estimates below based on the rotated factors (Table 10) the interpretation is across factors/criteria. These measure the variation of a variable in common with all the others together. For example, the values in the principal diagonal for import growth of.564 (Egypt) means that 56 percent of the growth in imports can be predicted from (and is dependent upon) data on the remaining eight indicators. By using commodity data on the eight indicators we could determine the degree of import growth for a commodity within 61 percent of the true value, on the average.

The factor matrices generated are presented in Table 10. The number of factors (columns) is the number of substantively meaningful independent (uncorrelated) patterns of relationships among the variables. The factors may be thought of as providing evidence for three different kinds of influence (causes) on the data, as presenting three categories by which these data may be classified, or as illuminating three empirically different concepts for describing the country’s commodity characteristics.

The loadings, coefficients (x), measure which indicators are involved in which factor pattern and to what degree. They can be interpreted like correlation coefficients, i.e. the square of the loading multiplied by 100 equals the percent variation that a variable has in common with a rotated pattern.

From the Table 10(i), if the pattern is limited to those indicators with at least 25 percent of their variation involved in a pattern (loading of.50, squared and multiplied by 100); then the first pattern of interrelationships (Factor I or F1) involves high growth in production (.91), growth in imports (.89), and moderate/low - production variability (.55). Similarly for Belize this would involve share in calories (.80), production variability (.75) and growth in Production (.56).

The column headed communality is the proportion of an indicator’s total variation that is involved in the factor patterns. The coefficient (communality) shown in this column, multiplied by 100, gives the percent of variation of a variable in common with each pattern. Communality is also a measure of uniqueness. By subtracting the percent of variation in common with the patterns from 100, the uniqueness of an indicator is determined. This indicates to what degree an indicator is unrelated to the others or to what degree the data on an indicator cannot be derived from (predicted from) the data on the other indicators. For example, self-sufficiency has a communality of.64 for Egypt (.89 for Belize). This implies that 64 percent (89 for Belize) of the self-sufficiency rate for commodities in Egypt (Belize) can be predicted from knowledge of commodity values on the three factors; and that 36 percent (11 for Belize) of it is unrelated to the other eight indicators.

The ratio of the sum of the values in the communality column to the number of variables, multiplied by 100, equals the percent of total variation in the data that is patterned. Thus it measures the order, uniformity, or regularity in the data. As can be seen from Table 10, for the nine indicators, the three patterns involve 47.6 (68.2 for Belize) percent of the variation in the data. That is, one can reproduce about 48 (68 for Belize) percent of the relative variation among the commodities in Egypt on these nine indicators using the commodity scores on the three factors.

The percent of total variance shows the percent of total variation among the indicators that is related to a factor pattern. This figure thus measures the relative variation among the commodities in the original data matrix that can be reproduced by a factor or pattern - it measures a factor’s comprehensiveness and strength. The percent of common variance indicates how regularity in the data is divided among the factor patterns.

The upper part of Table 10 (above the thick line) displays the factor pattern matrix and provides an interpretation of the factors. The values in this matrix are standardized regression coefficients, which are functionally related to the partial or semipartial correlation between an indicator and the factor when other factors are held constant. Therefore, a value in this matrix represents the individual and non-redundant contribution that each factor is making to predict a subset of indicators. The coefficients greater then.30 are used in the interpretation of the factors.

Thus from Table 10(i) for Egypt, we observe that the indicators significantly loaded on the first factor are: growth rate in production; growth rate in imports; and production variability. This indicates that commodities (see respective commodities in factor score table below) that have high growth in production are also those with high import growth rates and high to moderate production variability. The second factor is significantly loaded with the following indicators: import dependency and self-sufficiency. This implies that commodities on which Egypt relies for self-sufficiency are mostly imported. The third factor is loaded with indicators of share in production; share in calories; and area harvested implying that commodities that are produced domestically contribute significantly to calorie intake in Egypt. A similar interpretation can be arrived at for the Belize results. Here factors loaded significantly on the first factor indicate that commodities that contribute significantly to calories are those with high variability in production (negatively), with a high to modest share in total agricultural production; growth in production and growth rate in imports. Factor 3 here also includes share in calories as having significant loading as was found for factor 1, although the interpretation is different. Here, it indicates that products which contribute negatively to calories are those that have large area harvested but very prone to production displacement by cheap imports and with moderate-low share in total agricultural production. Given that in just these two country case studies, the second factor (loaded on import dependency and self-sufficiency ratio) for both countries and with a significantly high degree of correlation amongst them, either one can be used as an indicator of the other without any loss of information. Thus, our factors here can be classified as follows: Factor 1 - rural development; Factor 2 - livelihood Security; and Factor 3 - food security. Further, as indicators are loaded across each of the factors, and given the fact that the criteria/indicators are interlinked, our hypothesis that a commodity needs to satisfy only one criteria seems to be supported by these initial results. More case study results, coupled with a comprehensive framework for reducing the dimensionality of the indicators, are being developed to further investigate possible generalization.

TABLE 11 (i)
Egypt - Commodities with positive scores on each of the factors

TABLE 11 (ii)
Egypt - Commodities with positive scores on each of the factors

Table 11 presents the products with positive contribution (scores) for each of the factors. A considerable degree of overlap could be observed among the product scores that are making a relevant contribution across each of the factors. This further supports the view that a product needs only satisfy one criterion to qualify for Special Product status. Ranking of the products based on only positive factor scores also provides a mechanism to reduce the number of products.

In summary, Tables 10 and 11 together support the notion that a commodity needs to satisfy only one criterion for "Special Products" status. For example, Table 10 shows that there are no non-zero elements verifying that the indicators are related to all the criteria. Further, Table 10 supports this point by indicating that the majority of the commodities contribute to each factor (criteria). Moreover, given the distribution of the indicators by factors one can conclude that the factor, F1, F2 and F3 represent rural development, livelihood security and food security respectively. The same outcome is found for Egypt and Belize. The critical conclusion from this analysis of the indicators based on products of two countries is that it is possible for countries to select only a few (reduced number of) indicators based on their specific national situation to link products effectively to the criteria designated for "Special Products" selection; and that for products to qualify for Special Product status, they only need to fit at least one of the indicator/criteria.