Does export-market participation improve productivity? Evidence from Spanish manufacturing firms

Abstract

This article has a dual aim. First, it sets out to underline a learning-by-exporting effect in Spanish firms between 1991 and 2002. It further seeks to outline the conditions allowing firms to benefit from these spillover effects. Using a propensity score matching method, a group of firms having entered the export market (treatment group) is compared with a similar group of non-exporting firms (control group), and difference-in-differences regressions are carried out. The results show a cumulative productivity differential of 32% for the first four years of exporting, with continuous improvement in productivity. After three years of exporting, productivity gain is still approximately 10%. This study shows that increases in capacity utilisation and competitive pressure from foreign markets are insufficient to explain this causal link between exporting and total factor productivity (TFP). It is thus possible to deduce the presence of a learning-by-exporting effect, benefiting firms with sufficiently qualified employees and which are already engaged in international relations (due to foreign suppliers and/or foreign equity participation).

Keywords:

• exports
• productivity
• firm heterogeneity
• learning-by-exporting
• propensityscore matching

JEL Classifications:

• D21
• D21
• D24
• F14
• O33

Notes

 1. This literature is summarised by Wagner (2007) and Greenaway and Kneller (2007).

 2. We take into account only entering and non-exporting firms. Exporting firms are not considered here.

 3. We distinguish 17 Spanish locations (Andalucia, Aragon, Asturias, Baleares, Canarias, Cantabria, Castilla-la-Mancha, Castilla-Leon, Cataluna, C. Valenciana, Extremadura, Galicia, Madrid, Murcia, Navarra, Pais Vasco, La Rioja).

 4. 50 is the median size for the whole sample.

 5. The matching test is performed using the PSMATCH2 software in Stata provided by Leuven and Sianesi (2003).

 6. This represents firms entering the export market.

 7. The matching algorithms differ in the construction of weights.

 8. The import penetration rates are calculated by using COMTRADE database. Import penetration rate is measured as ratio imports/(imports + domestic output) at the 2-digit level. This indicator reflects the importance of foreign producers.

 9. Details regarding the cleaning process are provided in the Appendix 1.

10. Since there are fewer observations with zero-investment values in our sample than there would be in a sample of firms from developing countries, we are able to construct a consistent productivity measure.

11. In order to verify the robustness of our results, we also carried out the same analyses using the propensity score matching method with the nearest neighbour with replacement. The results turned out to be identical; they may be obtained from the authors upon request.

12. The balancing tests are explained in the Appendix 3.

13. The effect on TFP, expressed as a percentage, is obtained from the dummy variable as follows: [exp(coeff (dummy))−1] × 100.

14. Given that the foreign capital share coefficient is never significant, we have dropped this control variable in all regressions.

15. The fact that export destination plays no role in productivity differentials is not surprising given that, as is shown in Table 1, the vast majority of exports are produced for the European market. Indeed, among the firms that become exporters, only 14% do not sell on the European market. For the 86% that sell on the EU market, the mean of export share to EU countries is 85%.