In economics, the effects of tariff protection and tariff reform remain a contentious, fervently debated issue.1 In South Africa, decades of high tariff protection, of which domestic manufacturing sectors primarily benefited, are being followed by a period of substantial tariff reform. To this end, endorsement by the South African government of the Uruguay Round of the GATT, in 1994, has manifested itself in the phasing-in of lower tariffs on imported goods, at times at rates accelerated beyond the commitment to GATT (Swanepoel et al., 1997).2 In international economic theory, the principal of comparative advantage implies that barriers to trade, for example tariffs, lower the welfare of the protected nation (Mohr et al., 1995: 458).
Nevertheless, the 1950s and 1960s saw many developing countries, particularly in Africa, adopt import substitution as a development strategy. But in the period post World War 2, the outward-oriented East Asian economies rapidly outperformed the inwardly focused African countries, a record that sustained through the 1990s, albeit dented by the emerging market crisis of 1997. Indeed, the theoretical literature has increasingly supported the view that countries exhibiting small domestic markets will suffer constrained growth in a closed economy structure (see Ray, 1998). This view recognises the failings of the conventional international economics assumption that imported and domestic goods, in a given sector, are perfect substitutes. When, as Armington (1969: 159) argues, imports and domestic goods are not perfect substitutes in consumption or production, the effects of tariffs, and thus of either an import substitution strategy or subsequent tariff reduction programme, depend critically on the magnitudes of the substitution elasticities estimated (Naude et al., 1999: 42).3 Using economic models to evaluate changes in trade policy generally requires, amongst other things, the conversion of policy changes into price effects. Trade policy models use these price shifts to determine how the policy under review is expected to affect output, employment, trade flows, economic welfare and other variables of interest. The direction and magnitude of a trade policy change on individual variables depends on the size of the shock, as well as the behavioural relationships present in the economy (Gallaway et al., 2001: 1). When evaluating policy shifts in an economic model, these behavioural relationships largely take the form of elasticities, which reflect the responsiveness of one set of variables to a change in a second set. For example, trade liberalisation reduces the relative price of imported to domestic goods. This leads to substitution towards imported products, the extent to which is dependent on the degree of substitutability. Subsequently, a key relationship for model analysis is the degree of substitution between imported and domestically produced goods as the relative price of those two goods changes, i.e. the Armington elasticity. This elasticity estimate is derived from the assumption that preferences are well behaved over a weakly separable product category that comprises similar, but not identical products. Crucially, these products are considered imperfect substitutes due to their differing countries of origin (Armington, 1969: 159). In the context of this study, this implies that even though goods produced in South Africa and the rest of the world may fall within the same product category, they are not perfectly substitutable for each other. With this in mind, it is useful, often critical, to ascertain the degree of substitution between foreign and domestically produced goods.
In general, knowledge of elasticities is important for aggregate issues, such as changes in tariffs or taxes. Policy changes of these kinds will effect a country's trade balance, level of income, and employment, the magnitude of the effects depending on the elasticity magnitudes (McDaniel and Balistreri, 2001: 1). Thus, the Armington elasticity forms an essential component of modelling the effects of international trade policy. Furthermore, applied partial and general equilibrium models employed to examine trade policy are almost all sensitive to trade elasticities (McDaniel and Balistreri, 2001: 12). Indeed, the Armington elasticity is a key parameter determining the quantitative, and in some cases qualitative results that policy makers use.4 Set against this backdrop, our study attempts to estimate Armington elasticities, at the industry level, for South Africa.
At this stage it is pertinent to establish a framework of model and elasticity estimateassessment. To this end, the analysis that follows has three objectives. First, to determine asuitable method that will, depending on the time series characteristics of the data, allow for the extraction of both short- and long-run Armington elasticity estimates. Second, to apply the method to each of the forty-five sectors under review to estimate the short-run and, where possible, the long-run estimates. Third, in light of the elasticity estimates generated, analyse possible policy effects for South Africa.
The discussion proceeds as follows. Section 2 reviews the literature content of Paul Armington's (1969) original contribution to trade policy analysis, and continues with a synopsis of industry level Armington elasticities estimated over the subsequent three decades, in both the international and domestic context. Next, Section 3 provides the mathematical counterpart to the review of Armington's (1969) exposition, and derives the base equation to be used in estimation. Section 4 examines the four data series required for the estimation procedure, while Section 5 explains the econometric implementation of the method outlined in Section 3. The results obtained from the econometric processes applied are summarised in Section 6. These results are then compared to previous South African and United States industry estimates. In addition, Section 6 highlights problems encountered during estimation that may impair the reliability of results. This relates specifically to weaknesses in data availability, as well as the use of a single equation approach to estimation. Section 7 concludes the study, summarising key results obtained.