The decision to export: a panel data analysis for Spanish manufacturing

Abstract

This paper aims to empirically analyse the determinants of the decision to export using a sample of Spanish manufacturing firms during the 1990s. The data are drawn from the Encuesta sobre Estrategias Empresariales. A panel data probit model is used that is estimated using maximum-likelihood techniques. The results show that regional and local spillovers and firm characteristics such as age, size, productivity, corporation, foreign ownership, R&D and advertising intensity, have a positive and significant influence on the probability of exporting. Furthermore, public sector oriented sales have a significant and negative impact.

Acknowledgements

We gratefully acknowledge the financial support provided by Project SEC2002-03812 of the Spanish Ministry of Education and Culture.

Notes

 Most of the studies about the decision to export by Spanish firms have been focused on a limited range of variables (Moreno and Rodríguez (1998) look at foreign capital, Labeaga and Martínez-Ros (1994) at size and technological activities, Alonso and Donoso (2000) at policy instruments designed to incentive exports, or Barrios et al. (2001) at R&D and intrasector spillovers). To the authors’ knowledge the first attempt to control for a wider range of firm level variables is Merino's (1998) cross-sectional analysis.

 Huerta and Labeaga (1992), Moreno and Rodríguez (1998), Merino (1998) and Merino and Salas (2000), analyse the decision to export using cross-sectional models.

 See Dunne et al. (1989), Baily et al. (1992), Roberts (1996), Tybout (1996) or Liu and Tybout (1996).

 This is referred to as the self-selection hypothesis. See for example Ericson and Pakes (1995), Pakes and Ericson (1998) or Baldwin and Rafiquzzaman (1995). Delgado et al. (2002) find evidence supporting the self-selection hypothesis using the ESEE.

 This will involve that the first two years of the sample cannot be used in estimation, since two past observations are needed to introduce the growth rate of domestic sales lagged one period.

 The sampling procedure of the ESEE is the following. Firms with more than 200 employees are requested to participate and the participation rate is approximately 70% of the number of firms in the population. Firms between 10–200 employees are randomly sampled by industry and size strata, holding around 5% of the population. Firms with less than ten employees are excluded from the survey. The influence of the sampling scheme in export probabilities has been taken into account in the empirical analysis by inclusion of a variable that controls for big and small firms. Analogously, the spillover variables are built up taking account of that sampling scheme.

 Including the firms that failed during the period of analysis would involve modelling the probability of failing and would complicate substantially the analysis. However, this assumption is not innocuous as shown in Esteve et al. (2003), who, using a sample drawn from the ESEE, find that exporting firms are less likely to fail.

 Both in Roberts and Tybout (1997) and Bernard and Jensen (2003) firm size has a positive influence on the probability of exporting. Bernard and Jensen (2003) also obtain a positive coefficient for productivity. In Merino's (1998) cross-sectional analysis, both the variables that proxy for size and productivity have a positive and significant coefficient.

 However, in Merino (1998) the coefficient of the variable that captures the influence of labour force quality in the probability of exporting is positive and significant.

 In Merino (1998) the coefficient of the variable capturing foreign ownership is not significant.

 Merino (1998) does not find a relationship between advertising intensity and probability of exporting. Moreno and Rodríguez (1998) only find a positive and significant relationship between advertising intensity and probability of exporting for firms of more than five hundred workers.

 Bernard and Jensen (2003) introduce the same variables capturing spillovers but these are always non-significant or have a negative sign (contrary to expected).