Intra-Industry Trade in the Manufacturing Sector of South Africa

Abstract

Previous studies have attempted to examine the level of South Africa’s intra-industry trade. Given the serious limitations of the various indices like the Grubel-Lloyd index that were used in prior studies, this study uses regression analysis to determine the significance of intra-industry trade in South Africa’s manufacturing trade. South Africa’s manufacturing trade is reasonably explained by the world share and similarity of South Africa and its trading partner countries. Since the Helpman theorem is based mainly on trade with differentiated products and since intra-industry trade refers to trade in these product varieties, the responsiveness of South Africa’s bilateral manufacturing trade to these variables is sound evidence that intra-industry trade constitutes a significant proportion of South Africa’s manufacturing trade with the rest of the world.

Notes

1. The REM is consistent only if it fulfils the following orthogonality conditions. The use of REM requires , and μ_ij are independent of υ_ijt. Moreover, the explanatory variables have to be independent of the μ_ij and υ_ijt for all cross-sections and time periods. On the other hand, the FEM remains consistent and is not subjected to the above kinds of orthogonality conditions as long as it is free from endogeneity or errors in variables. On the other hand, the REM has the advantage of more efficiency as compared to the FEM. Egger (2002) says that in situations when the REM does not meet the orthogonality requirements, only the FEM is consistent since it wipes out all time invariant effects (μ_ij).

2. The derivation of the equation is found in Helpman and Krugman (1985:165-167) and Feenstra (2004:144-147).

3. Some of these limitations are given in Feenstra (2004). A myriad of costs (in money, time, and currency risk) involved in making transactions across borders are probably not reflected in aggregate price indexes. Another problem with using price indexes is that they are nearly always measured relative to an arbitrary base period. This makes it impossible to compare the “level” of prices in different places when the base period for each index differs.

4. Prais-Winsten transformation is carried out to correct for serial correlation in the data set. It transforms the dependent variable and independent variables to their differences by deducting ρ times the lagged value of each variable from the current value of the same variable. ρ is determined for each cross-sections by running ordinary least square on residuals from the original specification on lagged residuals.

5. The Brűlhart B measure calculated using the following formula:

6. If easily available, four-digit SIC (standard industrial classification) could have given us better estimates as this level of disaggregation appear to correspond closest to the concept of an industry.