Measuring the export capability of South African regions

An earlier version of this article was presented at the Biennial Conference of the Economic Society of South Africa, 10–12 September 2007, Johannesburg.

Abstract

Recent research has identified what determines local exports and what policies might make them grow. Regions with higher Gross Domestic Product per capita, faster population growth, higher levels of skills, greater export diversification and shorter distances to ports have experienced faster export growth. However, the results of regression models apply to a theoretical representative region and do not allow one to establish where policy interventions will be most effective. This article constructs an index to identify the regions in South Africa that can export manufactured goods. It draws on the literature of the determinants of exports for indicators of the capability (or potential) to export across 354 magisterial districts in 1996 and 2001. The results show a positive relationship between export capability and export performance. The article identifies a number of front-runner magisterial districts along with those of high capability but low performance that stand to benefit most from industrial policy interventions.

Keywords:

• Export capability
• export performance
• subnational regions
• South Africa

1. INTRODUCTION

Since 1996, export-led growth has been a key element of the South African Government's Growth, Employment and Redistribution strategy. Exports have been promoted through various supply-side measures and incentives, a programme of tariff reductions and reforms, and the relaxation of exchange controls (Coetzee & Naudé, 2004). Recent research into South Africa's export performance by Edwards and Alves (2006) has, however, shown that the success of these policies has been mixed. Although the growth in exports, and specifically manufactured exports, seems impressive at first glance, it has not been enough to generate an export-led boom. South African manufactured exports remain resource-based or concentrated in products with a declining share in world markets. Nevertheless, Edwards and Alves (2006) conclude that exporters are responsive to policies and economic environments that raise the profitability of export supply. In its 2007 National Industrial Policy Framework, the Department of Trade and Industry (2007) reiterates that South Africa's key industrialisation challenge is to grow and diversify its manufacturing exports.

The policies and economic environments that promote exports can be examined at two levels. The first is an economy-wide level. For example, exchange rate depreciations show a positive relationship with export performance as these depreciations raise the profitability of export supply (see Todani & Munyama, 2005). Similarly, tariff liberalisation leads to improved export performance by reducing intermediate input costs and lowering the incentive to produce for the domestic market (see Harding & Ratts⊘, 2005). Edwards and Alves (2006) also find that the availability of skills and infrastructure appears to be an important determinant of export growth. Skills and infrastructure are, however, not only relevant at the aggregate, economy-wide level but also inseparable from the location of the exporter. Thus the second level of analysis is to consider the geography of the ‘economic environments and policies that raise the profitability of export supply’ (Edwards & Alves, 2006:497).

The physical location of exporters is significant to the extent that the region determines the natural endowments available to and the distances faced by the exporter. Policy interventions in the forms of human capital formation and infrastructure investment are also place specific. The national industrial policy framework (Department of Trade and Industry, 2007) states that infrastructure is an important way to foster industrial clustering, which benefits firms through efficiencies and learning effects. Support for industrial infrastructure has thus far been limited to Industrial Development Zones and matching support for municipalities under the Critical Infrastructure Programme. The policy framework states that work is currently underway to identify specific areas and corridors in which high economic need coincides with good economic potential (Department of Trade and Industry, 2007:49). This article contributes by constructing an index to identify the regions in South Africa that have the capability to export manufactured goods.

Recent research by Matthee (2007) and Matthee and Naudé (2008) found that the regions in South Africa that have experienced faster export growth are those with higher Gross Deomestic Product (GDP) per capita, faster population growth, higher levels of skills, greater export diversification and shorter distances to ports. Such regression analysis is useful for identifying the determinants of local exports, and by extension the possible policy levers for export growth, but it does not allow one to establish in which places the policies are more likely to bear fruit. This article constructs an index to measure the capacity of regions in South Africa to export manufactured goods. The notion of an Export Capability Index is inspired by the United Nations Conference on Trade and Development (UNCTAD, 2007) Foreign Direct Investment Potential Index for countries and by Zietsman et al.'s (2006) Growth Potential Index for towns in the Western Cape. The index creates a summary measure of capability, linked to specific regions. The ranking and analysis of regions can aid policy-makers in their efforts to promote export-led local economic development and prospective exporters in their location decisions.

The literature explains the exports from a particular region as determined by comparative advantage or resource endowments or the size of the home market and the cost of transport. This study used 16 measures of these determinants of exports to compile an Export Capability Index for 354 magisterial districts for 1996 and 2001. The article is structured as follows. Section 2 briefly overviews the theories that explain why places trade, Section 3 explains the measures used and how the Export Capability Index was constructed, Section 4 presents the rankings of magisterial districts (the high-capability locations, the second-tier locations in the established agglomerations and the districts at the periphery that show export capability), Section 5 compares export capability and export performance, and identifies the front-runner magisterial districts along with those of high capability and low performance that stand to benefit most from industrial policy interventions. Section 6 concludes and makes recommendations.

2. THE DETERMINANTS OF EXPORTS

Traditional explanations of trade focus on countries and trade between industries, but are also suited to explain trade between subnational places such as regions or localities. Typically, the endowment of factors of production such as natural resources and workers' skills determines a place's capability to export. Different levels of technology and returns to scale at different locations may also influence it. In addition, trade barriers and transport costs are likely to dampen exports from a particular location. This section discusses how trade theory and the new economic geography explain the determinants of exports. These determinants may in turn serve as measures of a particular region's capability to export.

2.1 Trade theory

The theory of comparative advantage (the Ricardian model) explains trade in terms of differences in labour productivity (Krugman & Obstfeld, 2000). Specifically, it shows that a country has an incentive to trade, even if it can produce the relevant commodities more efficiently than its trading partner. Trade between the two countries can benefit both, if each produces and exports that commodity in which it has a comparative advantage. This comparative advantage shows in the ratio between the two countries' pre-trade prices, and this ratio reflects only the labour costs (Du Plessis et al., 1987).

Comparative advantage can be measured by examining current trade patterns. Balassa (1965) calculates the index of revealed comparative advantage (RCA) as:

where X _SA,j is, for example, South Africa's exports of product j, X _SA,tot is South Africa's total exports, X _W,j is the world's exports of product j, and X _W,tot is total exports in the world. RCA > 1 indicates that South Africa is relatively specialised in the production of that particular commodity.

However, the RCA is only one explanation of trade. The Hecksher−Ohlin model extends the Ricardian model by including differences in countries' resources (factor endowments) in the model (Krugman & Obstfeld, 2000). This model is also referred to as the factor-proportions or factor endowment model, as it involves the interplay between the proportions of production factors that are available in various countries and the proportions they use to produce commodities. A particular factor endowment is a determinant of exports. The Hecksher−Ohlin model states that the country with an abundance of labour will produce and export the labour intensive commodity. In the country with ample labour, capital is the scarce production factor, which makes the price of capital high relative to the price of labour. Less capital and more labour will be used to produce the relevant commodity cost-effectively, and the country becomes labour intensive. This commodity is exported to the country with the capital-intensive industry (Du Plessis et al., 1987; Armstrong & Taylor, 2000). The opposite is true for the country with an abundance of capital.

In a modified version of the Hecksher−Ohlin model, Wood and Berge (1997) further investigate why countries' compositions of exports differ. They replace the two factors of production, capital and labour, with land and skills, and argue that the production of agricultural and manufactured goods for export requires different land-to-skills ratios. Manufacturing requires less land and more skills than agricultural production does, and vice versa. Countries with a high proportion of skills compared with their amount of land have a comparative advantage in manufacturing, whereas countries with a high proportion of land compared with their available skills have a comparative advantage in agriculture.

2.2 The new economic geography

Until recently, international trade theory assumed away all elements that might make explicit consideration of the geography of exports possible. Such elements as distance, transport costs, market size, scale economies and the clustering of economic activity have only recently been incorporated into trade models. Important initial contributions to integrating regional science and international trade theory have been made by Krugman (1979, 1980, 1991), Venables (2001), Fujita et al. (2001) and Fujita and Krugman (2004) in the form of the new economic geography (NEG).

The NEG accounts for the determinants of exports by explaining why similar regions have different economic activities (Ottaviano & Puga, 1997). The goal of the NEG is to provide a picture of the spatial economy by explaining the interaction between the forces that shape the geographical structure of an economy (Fujita & Krugman, 2004). These are either centripetal forces (e.g. market size effects, vibrant labour markets, and cost savings from infrastructure and knowledge spillovers) that pull economic activity together or centrifugal forces (e.g. immobile factors of production, land rents and costs of congestion) that have the opposite effect (Armstrong & Taylor, 2000; Fujita & Krugman, 2004). The NEG models also incorporate transport costs, which play a major role in forming spatial balances and agglomerating or dispersing economic activities (Lopes, 2003).

The core-periphery model explains agglomeration and trade; that is, the geographical clustering of manufactured exports in a location. The model consists of two regions (Region 1 and Region 2), two production sectors (agriculture and manufacturing) and two types of labour (farmers and workers). Table 1 presents the assumptions that hold for this model.

Table 1: Assumptions of the core–periphery model

Agriculture	Manufacturing
Products are homogeneous	Products are differentiated (each firm produces a different variety)
Located in only in one region	Located in both regions
Constant returns to scale	Increasing returns to scale
Farmers are immobile (farmers are equally distributed throughout both regions)	Workers are mobile
Agricultural goods are moved without cost between regions	Positive transport costs (in ‘iceberg’ form) are incurred in moving manufactured goods between regions
Sources: Krugman (1991); Fujita and Krugman (2004).

The immobility of the farmers is considered to be the centrifugal force, as they consume both agricultural and manufactured products. The centripetal force is more complex and involves a process called circular or cumulative causation. Circular causation consists of backward and forward linkages. Backward linkages occur where workers choose to live near the areas of production, and forward linkages where producers choose to put up factories near the larger market. If for some reason a large number of firms are located near to each other in Region 1, then a wider range of products will be produced in this region. The workers (who are also consumers) in Region 1 have more products to choose from than those in Region 2. They receive a larger income, because of the increasing returns to scale achieved by the firms. No transport costs are incurred, as the products are produced locally. Region 1's higher wages act as an incentive for workers in Region 2 to migrate to Region 1. The market expands in Region 1 and becomes larger than the market in Region 2. This creates the so-called ‘home-market’ effect; that is, firms will export those products for which there is already a large domestic demand. More manufacturing firms locate in Region 1 because it is more profitable, and a greater number of varieties are produced than before. These different product varieties are then exported to Region 2 (Krugman, 1980, 1991; Armstrong & Taylor, 2000; Brakman et al., 2001; Fujita & Krugman, 2004).

Transport costs are the catalyst for the home-market effect. In the model, these costs are assumed to be zero within a region and positive between two regions. Brakman et al. (2001) broadly define transport costs as the various elements that hamper trade –such things as tariffs, language differences and cultural barriers as well as the actual costs incurred in moving goods from one place to another. In the model, these transport costs create a propensity for agglomeration. Internal economies of scale in manufacturing mean that producing more at a single plant will lower costs. The manufacturer will of course incur transport costs if he wishes to sell his output in the other region as well, and will thus try to choose a location that maximises the cost saving from large-scale production and minimises transport cost (Brakman et al., 2001; Fujita et al., 2001; Krugman, 1991). If transport costs are high, trade will not take place – exports and imports will be so expensive that only home production will be possible. Production will be spread out to be close to where demand is. If transport costs are low, there will also be no trade or agglomeration since the two regions will be ex ante identical and neither will have the forces, such as a thick labour market or inter-industry linkages, that reinforce agglomeration. Thus, it is in an intermediate range that transport costs matter for trade and agglomeration. Below this threshold level of transport costs, manufacturers choose a location with a large local demand – it will be large precisely where the majority of manufacturers choose to locate. The result is agglomeration at the core and trade with the periphery (Brakman et al., 2001; Fujita et al., 2001; Krugman, 1991).

In summary, the NEG predicts that exports from a particular locality will depend on the size of the local market, the diversity of intermediate inputs, the thickness of the labour market and the transport costs. These determinants of exports may also serve as measures of the capability of a particular locality to export.

3. CONSTRUCTION OF AN EXPORT CAPABILITY INDEX

The literature presented above makes it clear that there are many determinants of exports from a particular locality. In the case of South Africa, recent research by Matthee and Naudé (2008) found that higher GDP per capita, faster population growth, higher levels of skills, greater export diversification and shorter distances to ports were significant determinants of exports from 354 magisterial districts in 1996 and 2001. However, regression models that identify the determinants of exports are regressions to the mean and the results are for a representative region. The aim of this article is to identify the places where policy-makers can apply their knowledge of the determinants to increase exports. The determinants may thus be argued to serve as measures of a particular locality's capability to export. The magisterial districts with the higher levels of skills and shorter distances to ports will have the greater capability to export. To evaluate a place's export capability objectively a researcher needs a balanced set of multidimensional criteria – an export capability index.

The notion of an export capability index is inspired by UNCTAD's (2007) Foreign Direct Investment Potential Index for countries and by Zietsman et al.'s (2006) Growth Potential Index for towns in the Western Cape.

3.1 Data and measures

The construction of the index draws together measures of export capability, based on the theory discussed in Section 2, for which data are available on a subnational level. Data considerations influence the geographical level of analysis and the measures that can be used.

A magisterial district was used as the spatial unit of analysis since data were available for the period 1996−2001. Before 2000 there were 843 municipalities in South Africa. These were then reduced to 237 municipalities to constitute so-called wall-to-wall local government. This change in demarcation means that municipal data are not comparable over the extended period. At the lower level of the magisterial district, however, data were available for the whole period. In addition, the magisterial districts define the location of cities and towns, whereas the municipalities are governing bodies of larger areas.

Several sources were used to obtain data for the three types of measure described below. Information about subnational exports was available from the South African Revenue Service at postal code level. This was mapped onto the particular magisterial districts. The data for the indicators of export capability were sourced from the Regional Economic Explorer database compiled by Global Insight (Regional Economic Explorer, 2007). Questions of the reliability of subnational data restricted the analysis to the Census years of 1996 and 2001, for which comprehensive source data were available. The Regional Economic Explorer database is compiled from a range of primary sources, including not only the Census but also the 1995 and 2000 Income and Expenditure Surveys, the 1999 October Household Survey and the 2000 Labour Force Survey (Statistics South Africa, 1995, 1996, 1999, 2000a, 2000b, 2001).1

The indicators of export capability used to construct the index were:

• a measure of RCA,

• Hecksher–Ohlin–Wood–Berge measures, and

• NEG measures.

The RCA variable assessed the structure of production in magisterial districts. The gross value added in different manufacturing subsectors was matched with South Africa's RCA in international trade. The RCA was obtained from Pearson (2007). This was matched to gross value added in the following subsectors: other mining and quarrying; food, beverages and tobacco products; wood and wood products; fuel petroleum, chemical and rubber products; and other non-metallic mineral products. Magisterial districts with a greater contribution to national gross value-added totals in these subsectors were seen to have the capability to export since their current production structure corresponded with the RCA.

The Hecksher–Ohlin model's measures of export capability draw on the Wood–Berge approach discussed in Section 2, taking skills per labour and land per labour as indicators of export capability. The skills-per-labour measure was the number of people with education levels equal to or better than Grade 12, expressed as a percentage of the total population in a magisterial district. The land-per-labour measure was the size of the magisterial district in square kilometres, expressed as a percentage of the economically active population in the magisterial district. Higher levels of education were positively associated with export capability, but more land per labour was negatively associated with the capability to export manufactures.

The NEG measures included a number of indicators of export capability from NEG theory. The size of the home market was measured by population and population density. The economic size of the home market was measured in terms of GDP per capita and disposable income.

The diversity of intermediate inputs was measured by five variables: the Tress index, the level of concentration or diversification in an economy (0 represents a totally diversified economy, while a number closer to 100 indicates a high level of concentration); the location quotient of manufacturing, the comparative advantage of an economy (a magisterial economy has a location quotient larger than 1, or a comparative advantage in manufacturing when the share of manufacturing in the local economy is greater than the share of manufacturing in the national economy); formal employment in manufacturing; agriculture as a share of the economically active population; and mining as a share of the economically active population.

The thickness of the labour market was also derived from five variables, arranged in two groups: size of the labour market, measured by economically active population, population density, and urbanisation rate; and quality of human capital, measured by the proportion of the population with education levels equal to or greater than Grade 12, and the human development index.

The size of the labour market was measured by taking into account the economically active population, population density and urbanisation rate. The quality of the human capital was measured in terms of the proportion of the population with education levels equal to or greater than Grade 12, as well as the human development index. The thickness of the labour market was measured in terms of the formal employment in manufacturing as a percentage of the economically active population. The final NEG measure was transport cost. This was measured in terms of the shortest distance between the particular magisterial district and the main export hubs for manufactured goods: Johannesburg, Cape Town, Durban and Port Elizabeth.

Finally, the method of constructing the index was principle component analysis (PCA). Export capability is a latent variable that cannot be directly measured, but the analysis can identify groups of variables that measure this construct. PCA helped to screen the data, extract the factors, determine the communality (i.e. the proportion of the common variance present in a variable; Field, 2005:630) and calculate the factor scores. The factor scores were then used for further analysis. They were aggregated to form the export capability index and used to identify high-capability locations.

3.2 Principle component analysis

The analysis was carried out using SPSS 15, starting with the data for the 354 magisterial districts pooled for the periods 1996 and 2001. The first step in the analysis was to standardise the variables. Standardised z-scores were computed by the formula:

where x _ik is the raw value of variable k for the magisterial district I, x¯ _k is the mean value of variable k for all magisterial districts, and σ _k is the standard deviation of the variable k. The z-score of variable k has a mean value of 0 and a standard deviation of 1.

The second step was to screen the data using a correlation matrix. The analysis requires measures that correlate fairly well, but not perfectly. When all the variables listed above were used, the correlation matrix showed high correlations between the economically active population and the size of the population (r = 0.948); consequently, economically active population was dropped from the analysis.

The third step in the analysis was to determine whether PCA was appropriate. Table 2 presents the test statistics for the Kaiser–Meyer–Olkin measure and Bartlett's test for sphericity. Kaiser–Meyer–Olkin is a measure of sampling adequacy and represents the ratio of the squared correlation between variables to the squared partial correlation between variables. The statistics ranged between 0 and 1. A value close to 1 indicates that patterns of correlations are relatively compact and PCA should yield distinct and reliable factors. Field (2005) indicates that values between 0.7 and 0.8 are good, and here the Kaiser–Meyer–Olkin statistic is 0.738. Additional analysis of the anti-image correlation matrix showed that the diagonal elements were greater than 0.05 and the off-diagonal elements were small. Thus PCA was appropriate for these data.

Table 2: Kaiser–Meyer–Olkin measure and Bartlett's test

Measure	Value
Kaiser–Meyer–Olkin measure of sampling adequacy	0.738
Bartlett's test of sphericity
Approximate chi-square	8686.889
Degrees of freedom	105
Significance	0.000

Bartlett's measure tests the null hypothesis that the original correlation matrix is an identity matrix. The test was significant, which means that the R-matrix is not an identity matrix and that there are relationships between the variables that can be included in the analysis. Again, the conclusion was that PCA was appropriate.

The extraction of the factors identified three factors with eigenvalues greater than 1. Together the factors explained 58 per cent of the variance of the construct export capability. Table 3 presents the communalities before and after extraction.

Table 3: Communalities

	Initial	Extraction
Revealed comparative advantage	1.000	0.691
Skills per labour	1.000	0.417
Land per labour	1.000	0.497
Size of the population	1.000	0.754
Population density	1.000	0.351
GDP per capita	1.000	0.797
Disposable income per capita	1.000	0.684
Tress index	1.000	0.624
Location quotient for manufacturing	1.000	0.552
Agricultural employment % of total employment	1.000	0.611
Mining employment % of total employment	1.000	0.296
Manufacturing employment % of total employment	1.000	0.705
Urbanisation rate	1.000	0.474
Human development index	1.000	0.780
Distance from metros	1.000	0.492
Note: Extraction method: PCA.

The communalities represent the amount of variance in each variable that can be explained by the three factors that have been retained. For example, 70 per cent of the variance associated with manufacturing employment is shared variance.

Finally, the component matrix shows the factor loadings for the three factors identified through the PCA. It is clear from Table 4 that the extracted factors do not correspond neatly with the classification of the determinants of trade from theory. Factor 1 contains the measures of comparative advantage, skills-to-labour and many of those from the NEG. Capability to export is positively associated with a production structure that corresponds with the RCA and with higher levels of education. There is also a positive relationship with measures of the size of the home market, with local specialisation in manufacturing and manufacturing employment, and a negative relationship with distance. Squaring the factor loadings gives an estimate of the amount of variance in a factor accounted for by a variable. For example, skills per labour accounts for approximately 38 per cent of the variation. Factor 2 seems to represent indicators of a limited capability to produce manufactured exports. These include having a greater share of land per labour and more agricultural employment as a percentage of total employment. These coincide with smaller populations and magisterial districts that are less densely populated. Factor 3 captures South Africa's minerals economy that, like agriculture, seems to detract from the capability of a magisterial district to export manufactures. Where mining employment makes up a greater share of total employment, the local economy is more diversified and less likely to be specialised in manufacturing.

Table 4: Component matrix

	Component
	Factor 1	Factor 2	Factor 3
Human development index	0.810
Manufacturing employment % of total employment	0.799
Revealed comparative advantage	0.797
GDP per capita	0.765
Disposable income per capita	0.753
Skills per labour	0.613
Location quotient for manufacturing	0.568		−0.475
Distance from metros	−0.542
Urbanisation rate	0.533	0.434
Agricultural employment % of total employment		0.734
Size of the population	0.597	−0.622
Land per labour		0.524
Population density	0.405	−0.425
Tress index			0.623
Mining employment % of total employment			0.515
Note: Extraction method: PCA.

On the basis of the above analysis, the final step was to calculate factor scores for each of the magisterial districts. The analysis was repeated for 1996 and 2001 independently and the factor scores calculated. The scores of Factors 2 and 3 were then subtracted from the scores of Factor 1 to form a composite indicator of export capability. The following section presents the rankings of magisterial districts: the high-capability locations, the second-tier locations in the established agglomerations and the districts at the periphery that show capability to export manufactures.

4. RESULTS AND RANKINGS

Following the PCA explained in Section 3, an export capability index was compiled for the 354 magisterial districts in South Africa for 1996 and 2001. Figure 1 shows a scatter plot of the export capability index values and the log of manufactured exports per magisterial district in 1996. It shows that there is a positive relationship between the capability to export, as measured by the index, and manufactured exports. There are a few additional points to note. First, there were four key outliers with high export capability and significant exports in 1996. These were the four agglomerations around the cities of Durban, Johannesburg, Cape Town and Pretoria, which together produced 41 per cent of manufactured exports. Second, the majority of the magisterial districts had varying capability to export and varying exports. Third, there were a number of districts (160 to be exact) that did not export at all in 1996. Some of these districts had above average capability to export, but the mass lies below the national average.

Figure 1: Export capability index and manufactured exports in 1996

A box plot is a convenient way to assess the distribution of export capability across South Africa's magisterial districts. Figure 2 shows a normal, if somewhat thin-tailed, distribution of export capability, where most places have some potential and there are a few high-potential and limited-potential outliers. It also shows a number of changes over the period. Places like Umlazi, Pinetown and Umzinto emerge as contenders with export capability. For a more detailed discussion of the rankings, Table 5 presents the top 40 magisterial districts by export capability index in 1996, along with their percentage contribution to actual manufactured exports.

Figure 2: Box plots of export capability index

Table 5: Export capability index rankings

Magisterial district	Export potential index 1996	Manufactured exports as % of total in 1996	Magisterial district	Export potential index 2001	Manufactured exports as % of total in 2001
Durban	4.33	11.49	Soweto	5.11	0.12
Port Elizabeth	4.10	3.35	Mitchells Plain	4.54	0.00
Johannesburg	4.08	20.67	Port Elizabeth	4.35	4.50
Inanda	3.84	0.31	Inanda	4.26	0.57
Vanderbijlpark	3.71	0.63	Durban	4.22	9.43
Goodwood	3.66	0.78	Goodwood	4.07	0.76
Soweto	3.54	0.09	Alberton	3.96	0.59
Alberton	3.52	1.44	Johannesburg	3.87	13.98
Mitchells Plain	3.43	0.00	Vanderbijlpark	3.71	0.16
Pinetown	3.37	1.07	Pinetown	3.57	0.75
Pretoria	3.10	4.80	Umlazi	3.54	0.24
Germiston	3.10	3.35	Pretoria	3.4	11.04
Kempton Park	3.07	4.03	Kempton Park	3.31	3.17
Vereeniging	2.99	0.64	Wynberg	3.19	1.03
Pietermaritzburg	2.96	1.26	Benoni	3.16	0.49
Umzinto	2.96	1.21	Vereeniging	3.02	0.35
Benoni	2.91	0.45	Pietermaritzburg	2.99	1.80
Wynberg	2.83	1.00	Boksburg	2.94	1.01
Eshowe	2.83	0.05	Uitenhage	2.85	2.19
Uitenhage	2.75	0.54	Germiston	2.84	4.15
Boksburg	2.74	1.12	Umzinto	2.8	0.97
Lower Umfolozi	2.58	8.81	Wonderboom	2.63	2.00
Kuilsrivier	2.53	0.25	Eshowe	2.61	0.04
Lower Tugela	2.51	0.05	Roodepoort	2.52	0.25
Temba	2.47	0.05	Kuilsrivier	2.48	0.35
Krugersdorp	2.44	0.30	Krugersdorp	2.47	0.34
Bellville	2.43	1.26	Temba	2.46	0.03
Springs	2.41	0.62	Lower Tugela	2.4	0.05
Wonderboom	2.36	0.79	Springs	2.37	0.41
Malmesbury	2.28	0.04	Randburg	2.34	9.91
Umlazi	2.27	0.41	Malmesbury	2.33	0.08
Randburg	2.26	6.23	Bellville	2.32	1.12
Roodepoort	2.22	0.46	Lower Umfolozi	2.32	3.73
Sasolburg	2.21	0.39	Chatsworth	2.31	0.15
Estcourt	2.19	0.10	Brakpan	2.25	0.17
Mtunzini	2.18	0.10	Ga Rankuwa	2.16	0.00
Nigel	2.17	0.55	Nigel	2.15	0.08
Brakpan	2.15	0.23	Soshanguve	2.12	0.00
Newcastle	2.11	0.20	Sasolburg	2.05	0.46
Paarl	2.09	0.21	Newcastle	2.03	0.13

It is useful to consider the geography of the magisterial districts. The places with high export capability are not randomly scattered across South Africa but tend to cluster together. There are a number of ‘export clubs’ around the main port cities as well as in the economic agglomeration of Gauteng province. In the discussion that follows, the magisterial districts are grouped as high performers, the close neighbours in the established agglomerations, and those at the periphery with export capability.

Table 5 shows that Durban had the greatest export capability in the 1996 index ranking. Durban is South Africa's busiest sea port but contributed just 11.49 per cent of total manufactured exports in 1996. It has a number of neighbouring magisterial districts that also show significant export capability. Inanda and Pinetown were found high in the rankings – Pinetown contributed 1.07 per cent of South Africa's manufactured exports in 1996. Umzinto, which is also part of Durban's metropolitan area, known as eThekwini, ranked 16th in the export capability index and contributed 1.21 per cent of manufactured exports in 1996. On the southern side of this metro area was Umlazi, ranked 31st and contributing 0.41 per cent of these exports in 1996.

Port Elizabeth was ranked second in 1996 and contributed 3.35 per cent of manufactured exports. This should be considered together with nearby Uitenhage, which was ranked 20th and contributed 0.54 per cent of these exports.

In third position was Johannesburg, which also contributed the greatest share of manufactured exports in 1996, at 20.67 per cent. Johannesburg is not an isolated magisterial district but the centre of South Africa's main economic agglomeration. Along with Johannesburg, one should also consider the export capability and performance of 14 other magisterial districts: Germiston, Kempton Park, Boksburg, Springs, Benoni, Brakpan and Nigel to the east of Johannesburg; Roodepoort, Randburg and Krugersdorp to the west; and Alberton, Soweto, Vereniging and Vanderbijl to the south . Some of these magisterial districts also make large individual contributions to total manufactured exports, with Randburg contributing 6.23 per cent, Kempton Park 4.03 per cent, Germiston 3.35 per cent, Alberton 1.44 per cent and Boksburg 1.12 per cent in 1996. Together Johannesburg and its neighbours (within a 60 kilometre radius) contributed 40 per cent of South Africa's manufactured exports in 1996.

The third major agglomeration is the greater Cape Town metropolitan area. It is, however, limiting to consider only the Cape magisterial district, which is ranked 43rd in the 1996 capability index but accounted for 4.3 per cent of manufactured exports. The greater area also includes Goodwood, Mitchells Plain, Wynberg and Bellville, which all show export capability in the top 40 magisterial districts. Within a 45-kilometre range, Kuilsrivier, Paarl and Malmesbury are also found high in the rankings. Together these magisterial districts contributed 3.95 per cent of South Africa's manufactured exports in 1996.

The last of the high performing areas was Pretoria, ranked 11th by export capability in 1996 and contributing 4.8 per cent to total manufactured exports. Whether Pretoria is in fact a separate cluster may be point of contention. It could be argued that it forms the northern part of the greater Gauteng cluster around Johannesburg. In fact, the Pretoria magisterial district is only 50 kilometres from Johannesburg, with the Randburg magisterial district in between. To the north of Pretoria there are two neighbouring magisterial districts that are also high in the export capability rankings and may be considered part of the cluster: Wonderboom, 35 kilometres to the north of Pretoria, and Brits and Temba, which are another 28 kilometres further.

Aside from the high-capability locations and their established agglomerations, Table 5 includes a number of magisterial districts at the periphery that also show high export capability for 1996. These districts are clustered in KwaZulu-Natal Province and include Eshowe (ranked 19th), Lower Umfolozi (ranked 22nd), Lower Tugela (ranked 24th) and Mtunzini (ranked 36th), which lie between Durban and Richards Bay harbour. To the northwest of the province are Escort (ranked 35th) and Newcastle (ranked 39th), which together contributed 0.3 per cent to total manufactured exports in 1996.

Over the period 1996−2001, the top 40 magisterial districts remained consistently high in the export capability rankings. Table 5 highlights some changes but few new entrants. The notable new entrants are again situated close to the major agglomerations; for example, Chatsworth near Durban and Ga Rankuwa and Soshanguve near Pretoria.

In summary, this section has shown that the export capability index provides a reasonable description of the geographical clustering of exporters in South Africa. The top 40 ranked magisterial districts produced 79 per cent of the total manufactured exports in 1996 and 77 per cent in 2001. The exports originate mainly from Gauteng, Durban and Cape Town, and the close neighbours in the established agglomerations make an important contribution. However, to determine which magisterial districts present the best opportunities for local policy-makers and prospective exporters, the section that follows compares export capability and export performance in 2001.

5. COMPARING CAPABILITY AND PERFORMANCE

The previous section presented the magisterial districts with high export capability, but these are to some extent the ‘usual suspects’ of subnational economic activity in South Africa. The established agglomerations and their close neighbours dominate the rankings, but it is not clear which are the relative front-runners and which the under-performers. As a final step, it is possible to compare export capability and performance in 2001.

For such a comparison, a measure of export performance has to be calculated and compared with the export capability index. This approach again follows the method used by UNCTAD (2007) to calculate an inward Foreign Direct Investment performance index. In this article the magisterial districts are ranked by their exports relative to their economic size. The performance index is calculated by the formula:

where EXP _i is the export performance index of the ith magisterial district, Export _i is the value of the manufactured exports of magisterial district i, and GDP _i is the value of its gross domestic product. Magisterial districts can then be ranked according to their export performance and this can be compared with their export capability. When the two indices are compared, it is possible to draw up a fourfold matrix of export performance and capability. Magisterial districts with high export capability and performance are labelled as front-runners. Then there are those with low capability but high performance and those with high capability but low performance, both of which should be of particular interest to policy-makers. Finally there are those with low capability and low performance. Table 6 presents this classification of magisterial districts. The cut-off points for the different groupings were the median export capability and performance index values.

Table 6: Comparison of performance and potential in 2001

	High export performance	Low export performance
High export capability	Alberton, Bellville, Benoni, Boksburg, Brakpan, Bronkhorstspruit, Camperdown, Cape, Chatsworth, Durban, East London, Eshowe, Germiston, Glencoe, Goodwood, Heidelberg, Hermanus, Humansdorp, Inanda, Johannesburg, Kempton Park, King Williams Town, Krugersdorp, Kuilsrivier, Lions River, Lower Tugela, Lower Umfolozi, Malmesbury, Mokerong, Montagu, Mossel Bay, Mthonjaneni, Mtunzini, Newcastle, Nigel, Paarl, Pietermaritzburg, Pinetown, Port Elizabeth, Potchefstroom, Pretoria, Randburg, Randfontein, Roodepoort, Sasolburg, Simonstown, Somerset West, Soweto, Springs, Stellenbosch, Strand, Temba, Tulbagh, Uitenhage, Umbumbulu, Umlazi, Umzinto, Vanderbijlpark, Vereniging, Vredenburg, Wellington, Wonderboom, Wynberg, Zwelitsha	Alfred, Bethal, Estcourt, George, Kliprivier, Lichtenburg, Mitchells Plain, Ndwendwe, New Hanover, Nsikazi, Port Shepstone, Swellendam, Umvoti
Low export capability	Alexandria, Barberton, Bethlehem, Brits, Caledon, Harrismith, Hennenman, Madikwe, Nelspruit, Oudtshoorn, Robertson, Somerset East, Witbank, Worcester	Balfour, Bloemfontein, Bothaville, Bredasdorp, Christiana, Danhauser, Delmas, Dundee, Ermelo, Fort Beaufort, Groblersdal, Hankey, Heilbron, Highveld Ridge, Hlabisa, Knysna, Kroonstad, Mapulaneng, Middelburg, Mooi River, Mount Currie, Parys, Paulpietersburg, Piet Retief, Piketberg, Potgietersrus, Queenstown, Ritavi, Standerton, Stutterheim, Umzimkulu, Ventersdorp, Volksrust, Vuwani, Warmbad, Waterberg, Witsieshoek

Table 6 shows a number of low-capability high-performance magisterial districts. The numbers in brackets in the following give the rankings for performance and potential. The magisterial districts present a varied profile. A number are close to major agglomerations and would benefit from proximity and spillovers. Alexandria (103rd, 63rd) is part of the Johannesburg metro. Brits (122nd, 20th) lies to the north of Pretoria, a neighbour to high performers such as Wonderboom and Temba. To the east, along the Maputu development corridor, there are Witbank (137th, 35th) and Barberton (133rd, 68th). Nelspruit (144th, 34th), the provincial capital of Mpumalanga, lies at the eastern end of the corridor and also showed good export performance in 2001. High-performers that rank low in the export capability index in the Western Cape are Worcester (97th, 8th), Caledon (128th, 59th) and Robertson (144th, 18th). The other magisterial districts are further away from the metropolitan areas. Oudtshoorn (177th, 19th) is in the southern Cape, but still close to George (a high-capability area) and Mossel Bay (a front-runner magisterial district). Similarly, Somerset East (145th, 69th) is close to the port of East London. Bethlehem (129th, 83rd), Harrismith (92nd, 47th) and Hennenman (98th, 66th) are three magisterial districts that show low export capability but low performance in the northern Free State.

Where the magisterial districts above show high export performance despite low capability, there are also several with high capability according to the index but little or no exports. Of these 13, eight are in KwaZulul-Natal province, in addition to the high-potential high-performance magisterial districts already identified there: Estcourt (48th, 95th), Kliprivier (44th, 97th), Alfred (66th, 107th), Port Shepstone (42nd, 113th), Ndwendwe (76th, 120th), New Hanover (72nd, 98th), Nsikazi (55th, 123rd) and Umvoti (58th, 181st). In the Western Cape, Mitchells Plain (2nd, 150th) is part of the greater Cape Town metro and was ranked second in terms of export capability in 2001, and presents a key opportunity in the province. Swellendam (87th, 126th) is also close by. Further afield in the southern Cape, George (73rd, 94th) also showed high capability in 2001.

6. CONCLUSIONS AND RECOMMENDATIONS

This article has argued that one of the key challenges facing South Africa is to increase and diversify its manufacturing exports. For this to happen, the location of exporters is significant. Recent research found that the regions in South Africa that have experienced faster export growth are those that have a higher GDP per capita, faster population growth, higher levels of skills, greater export diversification and shorter distances to ports (Matthee & Naudé, 2008). The aim of this article has been not to identify policy levers but to identify the places where policies are more likely to bear fruit. Policy interventions in the form of human capital formation and infrastructure investment are place specific, and government is currently identifying specific areas and corridors for future investment initiatives.

The study this article is based on constructed an index to identify the regions in South Africa that have the capability to export manufactured goods. The analysis used PCA. Export capability is a latent variable that cannot be directly measured, but the analysis can identify groups of variables that measure the construct. PCA helps to screen the data, extract the factors, determine the communality and calculate the factor scores, which were then aggregated to form the export capability index and used to identify high-potential locations.

The results showed that the export capability index provides a reasonable description of the geographical clustering of exporters in South Africa. The top 40 ranked magisterial districts produced 79 per cent of total manufactured exports in 1996 and 77 per cent in 2001. The exports originate mainly from Gauteng, Durban and Cape Town, and the close neighbours in the established agglomerations make an important contribution.

To further analyse which are the relative front-runners and which the under-performers, the export capability was compared with export performance in 2001. A number of magisterial districts with low capability but high performance and others with high capability but low performance were identified, both of which should be of particular interest to policy-makers.

Expanding the manufacturing sector to its full potential is a focus of the Department of Trade and Industry. In a recent statement, Minister Mpahlwa (cited by Van der Merwe, 2008) states that interventions are required to enhance South Africa's industrial development, and that those interventions should be ‘pursued vigorously’. This paper therefore serves as such an intervention by identifying the places where policies are most likely to bear fruit.

The recommendation for further research is for more detailed profiles and case studies of the high-potential low-performance and the low-potential high-performance magisterial districts. Great care should be taken when government tries to pick winners, and the locations of investment and development initiatives should be scrutinised by researchers and policy-makers alike.

The authors would like to thank two anonymous referees for comments that greatly improved the article. All errors and omissions remain their own.

Notes

An earlier version of this article was presented at the Biennial Conference of the Economic Society of South Africa, 10–12 September 2007, Johannesburg.

¹In-depth analysis of the compilation of subnational data falls outside the scope of the article. Source documentation for the REX database describes the compilation of the data in some detail. Suffice to say that the data related to the Census years are as reliable as one is likely to find. An earlier (draft) version of the paper also reported 2006 estimates of the export capability index, but an anonymous referee pointed out that one then becomes subject to the uncertainties of out-of-sample forecasts. Consequently the analysis was restricted to 1996 and 2001. The results for 2006 are available on request from the authors.