Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Table 2.3 Key to names of 34 economic and social variables in Sub-Saharan Africa by 19 countries (mainly in the late 1980s)

Variable names in programme:

Names of countries in the programme

Sources: World Bank (1991a, 1990, 1989b); UNDP (1991); FAO (1991a,b,c, 1992).
Note: The "Africa" of this study is "Africa south of the Sahara." By convention it excludes South Africa. No doubt this practice will now change, but I have retained it in order to use standard data sets with that name. Africa south of the Sahara did include Botswana, Lesotho, Namibia, and Swaziland, but their trade figures are included with those of South Africa. So, lacking vital data, I excluded them, but they will be included in some of the total data analysis for SS Africa used, rather than spend time in calculating the sums for the countries available. Angola was also excluded from the list of countries because the data for 1989 or nearest years were simply inadequate.

The analysis included a correlation matrix (relationships) and tables of principal components (groups) (tables 2.4-2.6). Coefficients above.4000 have been identified as probably having some significance in indicating relationship possibilities, although some of the variables, particularly those with a financial base such as the valueadded in agriculture, admittedly have somewhat skewed distributions.

The percentage of rural population below the poverty line (RUPO -table 2.4) correlated most strongly and negatively with energy consumption per capita (ENCA), international trade per capita (EXPC & IMPC), food supply per capita (CALC), value-added in agriculture (VAAG), food import dependency ratio (food imports to food available for internal distribution - FIMD), and GNP per capita (GNPC). There seem to be some positive relationship to arable land as a percentage of total land (ARLA) and weak negative relationships to the index of change in food production (FPIN) and population growth (POGR), but only a very weak or no correlation with the infant death rate per 1,000 live births (FIVE - sometimes regarded as an indicator of poverty), life expectancy at birth (LIFE), the percentage of adult illiteracy (ILLI), and the percentage share of exports in '`other" primary commodities (EXOP - mainly agricultural). Obviously in some countries the rural poor will survive better than in others, possibly owing to better access to food. A high percentage of rural people below the poverty line in the late 1980s seems for this group of SS African countries to be mainly associated with a low general level of energy consumption and of international trade and a low value in agricultural production (although little association, whether negative or otherwise, seems indicated with the use of modern inputs such as fertilizers, pesticides and tractors). The percentage of the labour force in agriculture (LABA - table 2.4) has only a low positive correlation with the percentage of the rural population below the poverty line (RUPO), but like the latter has negative relationships to energy consumption (ENCA), trade (EXPC & IMPC), valueadded in agriculture (VAAG), GNP (GNPC), and food import dependency (FIMD). It is positively related to the rural percentage of total population (RURP) and also to the rate of urban population growth (URBG - i.e. a high percentage of agricultural labour, but not necessarily of rural poor, seems likely to be linked with urbanward migration), the infant death rate per 1,000 live births (FIVE), and hectares per tractor (HATR), and negatively to the index of change in food production since 1979-1981 (FPIN). It also has a negative correlation with the average annual rate of inflation (are the governments of countries with a high percentage of agricultural labour less inclined to overspend?), but only a weak correlation with the agricultural population per hectare of arable land (POAR). The agricultural population per hectare of arable land (POAR - table 2.4) is itself related to very few of these variables - mainly to the use of fertilizers (FERT) and pesticides (HAPE - associated with the intensive development of agriculture), to urban growth (URBG), and to the share of merchandise imports in machinery and transport (IMPM).

Table 2.4 Twelve key variables from the correlation matrix

Table 2.5 Eigenvalues of the correlation matrix for the first 10 principal components

In order to examine the way in which these variables group together, principal components have been identified. From a table of eigenvalues of the correlation matrix (table 2.5) it appears that 25 per cent of the values are in the first component and nearly 90 per cent in the first 10 (PRINC1-10), but even by the sixth component the level of so-called "explanation" has fallen to only 7 per cent. Three of the components identified seemed particularly relevant for this study: the first, the seventh, and the eighth (PRINC1, PRINC7, and PRINC8 in table 2.6).

The first principal component (PRINC1) has highest values on energy consumption, GNP, value-added in agriculture and manufacture, international trade, and life expectancy. Negative values link to rural poverty, deaths of infants, and rural population percentage. This would appear to be a component of development, but with no particular link to general GDP growth or agricultural growth.

Table 2.6 Eigenvectors for the first 10 principal components

The seventh principal component (PRINC7) has a highest eigenvector value for fuel imports, followed by food consumption and population growth. It is also linked positively to the index of food production per capita and to a lesser extent to the export of minerals, metals, and fuels, but negatively to food import dependence and the percentage of rural population below the poverty line. This component would seem to be identified with a satisfactory level of local food production and self-sufficiency except for fuels paid for largely by minerals exports (the eigenvector for value-added in manufacture is negative).

The percentage of rural population below the poverty line has its strongest link (positive) to the eighth principal component (PRINC8), of which it is the second largest eigenvector, the largest being the percentage of arable land. Component 8 is also linked to population growth, inflation, and debt percentage of GNP, and is low on imports of energy and of machinery and transport, and on value-added in agriculture. It would seem that this component represents rural poverty linked to arable production coupled with poor economic performance and financial problems.

However, the relationships described are only weakly developed in components 7 and 8, which have levels of "explanation" of only 7 per cent and 5 per cent respectively. The principal component analysis has had only a limited value, partly because of its complexity in grouping the variables and partly because the more relevant components had only low "explanatory" levels. The more significant components from this data set, apart from the first, reflected mainly urban growth, well-developed commercial agriculture, aid, and trade, and tended to suggest relationships of interest in a context outside the scope of this paper. However, the evidence does point to a link between rural poverty and a poor overall economic performance linked to dependence on agriculture, particularly non-commercial agriculture. Although this may suggest that improved productivity should reduce rural poverty, the marked income disparities shown in tab]e 2.2 also suggest that attempts to increase wealth are likely to do more for the rich than for the poor unless linked to policies that are focused on improved income distribution.

Contents - Previous - Next