Competition between Brazil and other exporting countries in the US import market: a new extension of constant-market-shares analysis

Abstract

This article develops a new extension of the constant-market-shares model, attributing the gains or losses of market share of an exporter in a specific market to its competitors. The method is then applied to Brazil's exports of manufactured products to the US market, determining from which countries and by how much Brazil gained market share, and to which countries Brazil lost market share in the period between 1992, 1999 and 2004. The bilateral gains and losses of Brazil to OECD countries are shown to be related to changes in the relative unit labour costs of these countries through a two-period panel data analysis.

Acknowledgements

Many thanks for helpful suggestions go to my colleagues Getúlio Borges da Silveira and Hugo Boff.

Notes

¹ Tyszynski (1951) is one of the earliest applications of the method.

² The analysis may be extended to include several destination markets.

³ This presentation using vector notation follows that of Fagerberg and Sollie (1987).

⁴ See Leamer and Stern (1970) for a detailed and critical analysis of the CMS model. Their version of the model is slightly different from the one presented here, since they focus on changes in export revenue rather than on change in market share. As a result, a demand effect appears in their version. But if the demand effect is subtracted from the change in export revenues, the result is the difference between actual export revenue at the end of the period and the value that would have been necessary to maintain the macro share of the exporting country constant. This, in turn, is equal to the change in market shares times the size of the import market at the final year. That is the left hand side of Identity 3.

⁵ Furthermore, Richardson (1971) also pointed out correctly that if the market composition effect is added to the analysis, the order in which the product and market composition effects are calculated matters for the results. It is also well known that the effects depend on the level of product aggregation. See Bowen and Pelzman (1984) for a sensitivity analysis of the CMS model to changes in the base year, level of commodity aggregation, and definition of world market.

⁶ This extension of the model has been developed by Fagerberg and Sollie (1987).

⁷ A dynamic variant of this version of the CMS method has been applied by Wilson (2000) to compare the performance of one exporting country to a group of benchmark countries in a specific sector for every year in a period of several years.

⁸ See, for example, Chami Batista and Azevedo (2002), and Moreira (2004).

⁹ The data for US imports by country of origin used in this work are from the United States International Trade Commission (USITC) and are customs value (FOB) classified at the 5-digit level of the Standard International Trade Classification (SITC), Revision 3.

¹⁰ Throughout this article, manufactured products are SITC products from group 5 to 8.

¹¹ Since the competitiveness effect in the period 1992 to 1999 is being calculated by difference, the changes in market share are, in fact, being multiplied by US imports of each good in 2004. This is convenient for comparison with the other periods, though the results can be quite different from those calculated on the basis of 1999 US imports.

¹² If all products had been included (SITC from 0 to 9), the product effects would have been negative in both sub-periods, revealing that Brazil tends to export primary products whose demand are less dynamic than the US import market as a whole. Chami Batista (2004) compares the product effects of the largest Latin American exporting countries to the US in the period between 1996 and 2002 and shows that the slow growth of exports of non-differentiated resource-based products account for the negative product effects observed in most of these countries.

¹³ Products were grouped according to OECD classification of high-tech products, see Hatzichronoglou (1997).

¹⁴ SITC 79230 and 79240.

¹⁵ Twenty five OECD countries were considered: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, UK.

¹⁶ Canada, France, Germany, Italy, Japan, and the UK.

¹⁷ Calculated by difference.

¹⁸ See Chami Batista (2006) for an analysis of North-American and Japanese FDI abroad.

¹⁹ The data and methodology for computing relative unit labour costs can be found at http://www.oecd.org/eco/sources-and-methods.

²⁰ I have restricted the analysis to OECD countries, in view of the greater quality and consistency of the methodology applied in these countries to the series of relative unit labour costs. There is also evidence that the response of export demand (or the elasticity of substitution) to changes in the real and effective exchange rate with respect to OECD countries is different from the response to non-OECD countries, see Spilimbergo and Vamvakidis (2003).

²¹ Carlin et al. (2001).

²² Ricardian comparative cost advantage is the major theoretical inspiration for using relative unit labour costs as a determinant of export market shares.

²³ Fixed effects assume that the unobserved variable is correlated with our explanatory variable, while the random effects assume just the opposite. See Wooldridge (2003).

²⁴ Iceland was excluded.

²⁵ Iceland, New Zealand, Luxembourg, Greece, Norway and Portugal were excluded. Alternatively, Finland, instead of Portugal, was excluded from the sample with no significant changes. Brazil's gains and losses of nonhigh-tech products from and to these six countries were, in absolute value, the smallest of all OECD countries, considering the two sub-periods. Indeed, these countries accounted for about 1% of Brazil's gains and losses (sum of absolute values) to OECD in the two sub-periods together. Their exports to the US were also the smallest among OECD countries in 2004. Alternative sample sizes between 24 and 19 countries were also tested, but they did not add any relevant information.

²⁶ Differences in the rates of technological progress and in institutions tend to matter independently of costs, see Carlin et al . (2001).

²⁷ The following averages and lags were tried: RULC_1_1, 1_0, 2_1, 2_0, 3_1, 3_0 and 4_0.

²⁸ An increase in the RULC of a competitor is expected to raise Brazil's gains from that competitor, hence the sign of the coefficient is expected to be positive.

²⁹ Brazil's loss (gain) to Mexico is greater (smaller) than what the growth-rate of Mexico's relative unit labour cost would predict.

³⁰ The coefficients of the dummies for other countries were not significant either at this level of significance.

³¹ All products classified in this group (SITC 6) are also nonhigh-tech products.

³² Indeed, this group of products accounted for 48% of Brazil's gross loss to Mexico in the period 1992 to 2004, of which 73% were products of the steel industry.

³³ The coefficients of the dummies for Ireland in both sub-periods and for Korea in the first sub-period are also significant, though negative for Ireland and positive for Korea. The gains of Ireland from Brazil are concentrated in syringes, needles, catheters, cannulae and the like (SITC 87221), whilst the losses of Korea to Brazil in 1992 to 99 are almost exclusively in footwear with outer soles of leather (SITC 85148). Brazil's losses to Mexico in this group of products are somewhat concentrated in articles of apparel and clothing accessories (SITC (84).