Why don't firms export more? Product quality and Colombian plants

Abstract

Exporting firms around the world ship only a small fraction of their output overseas. For firms in a large country, such as the United States, this behavior can be explained by the existence of a large domestic market. For firms in a small lower income country, such as Colombia, the lower share of exports remains a puzzle. This paper begins by illustrating the failure of current models to explain plant export patterns in Colombia. Even models that do well in describing the US export distribution fail when confronted with the Colombian data. In response to this puzzle, this paper suggests that Colombia's export distribution can be explained with a two-dimensional productivity space where output productivity is considered separately from quality productivity. Predictions of this theory are tested on Colombian plant level data from 1981–1991. Overall, product quality is shown to be a significant factor in explaining the tendency for Colombian plants to under-export manufactured goods to the United States.

Introduction

Analyses of export sales in manufacturing surveys have established a puzzling phenomenon: most exporting firms sell only a small fraction of their output overseas.¹ For firms in a large country, such as the United States, this behavior can be explained by the existence of a large domestic market. In a small country, such as Colombia, the low export intensity of plants remains a puzzle. For example, among the 10–20% of Colombian manufacturing plants that exported during the 1980s, the average export share was roughly 20%.² The distribution of export shares for Colombia, in Fig. 1, continued during the export boom of the 1980s. This phenomenon existed regardless of plant size, since the larger plants also had low export shares, as reported in Table 1.

These stylized facts are inconsistent with virtually any model of trade that incorporates varieties or firms; particularly considering that Colombia's market size was less than one percent of its primary export destination, the United States. The puzzle is amplified by the existence of sunk costs to exporting: after investing the time and energy to export to the U.S. market, it seems sub-optimal to sell small quantities to such a large market.³ An added complication is explaining the plants that do export at large intensities. Within industries, there are a few very high-intensity exporters competing with the many low-intensity exporters.

This paper has two main parts. The first part contains stylized facts about the export intensity of Colombian manufacturing plants. It also contains results showing that the Bernard et al. (2003) model – hereafter referred to as the BEJK model – cannot explain these facts. In the second part, we show that low product quality provides a significant explanation for why Colombian plants do not export more. More precisely, this paper provides evidence consistent with two-dimensional productivity: the productivity to create output and the productivity to create a quality product.

Furthermore, this paper adds an additional dimension to recent works on the interaction between within-industry heterogeneity and the decision to export or undertake foreign investment. From an empirical perspective, the sunk cost to exporting has been documented for Colombian manufacturing plants in Roberts and Tybout (1997b) and for the United States companies by Bernard and Jensen (2004b). Productivity-based self-selection into export markets has been demonstrated by Clerides et al. (1998) and Bernard and Jensen (1995), among others. Aw et al. (2000), on the other hand, finds evidence of self-selection in the Taiwanese market, but not in Korea.

Building upon the theoretical dynamic industry model of Hopenhayn (1992), Melitz (2003) offers a model of trade in which only the most productive firms export. Along a similar vein, BEJK propose a Ricardian model of international trade in which the most productive producer (after incorporating trade costs) supplies the product to each country.

The evidence in this paper suggests that the one-dimensional approach to productivity is insufficient to explain low-intensity exporters in Colombia. Instead, we propose analyzing the productivity of quality separately from the productivity of output. This approach of using two-dimensional productivity has been applied to wage inequality issues in Manasse and Turrini (2001), who use fixed firm effects to determine productivity levels. Verhoogen (2004) adds another dimension by allowing Mexican firms to choose their output qualities. The results in this paper are consistent with a model that draws productivities with two dimensions, quality and output.

This microeconomic understanding of product quality adds an additional degree of understanding to what we know happens at the aggregate level. For example, we describe an additional reason why inter-industry trade is greater among countries with similar income levels. This complements existing empirical work on the Linder (1961) theory.⁴ Likewise, we describe one example at the micro-economic level of how non-homothetic preferences can affect the export patterns of developing countries.⁵ Our results also add a new level of microeconomic detail to the conclusion of Hummels and Klenow (2005) that “within categories, richer countries export more units at higher prices to a given market, consistent with producing higher quality”. In the case of a small and less wealthy economy like Colombia's, we find that relative price gaps matter more when exporting to wealthier countries in industries with more varieties.

The analysis proceeds as follows. Part 2 describes the data. Part 3 clarifies the stylized facts. Part 4 posits that separating quality productivity and output productivity could explain the small export shares. In part 5 a static prediction of the theory is tested: that plants should have larger export shares when they are producing qualities more comparable to those of the United States. Part 6 concludes.