Learning by exporting or self-selection into exporting?

ABSTRACT

This paper explores the export-productivity relationship using firm-level data from Egypt over the 2003–2008 period. Previous studies using data from developed countries suggest that self-selection is the main driver of the exporter premium. Using a propensity-score matching difference-in-difference approach, we find that both labor productivity and total factor productivity are significantly higher for exporters than for non-exporters. On average, labor productivity and total factor productivity are, respectively, 43% and 61% higher for exporting firms than for domestically-oriented firms. Accounting for the level of development of destination countries, we find that this export premium is due to a learning-by-exporting process rather than just a self-selection of more productive firms into exporting. In contrast to exporters to OECD countries, exporters to Non-OECD countries self-select into export markets, signaling the importance of the technical assistance from OECD buyers.

KEYWORDS:

• Exports
• self-selection
• learning from exporting
• productivity

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See Wagner (2007a) for a recent survey of the firm-level evidence on this issue.

2 Sunk start-up costs include any extra costs of locating foreign demand and competition, establishing distribution networks, adjusting product characteristics to meet foreign tastes, regulations and standards.

3 This argument is in line with work on learning-by-doing such as Arrow (1962).

4 The ‘El-Infitah’ or ‘Opening’.

5 Exporters may also be more efficient by reducing X-inefficiency, while using the same technology (Nishimizu & Robinson, 1984).

6 Domestically-oriented firms are often protected from international competition, especially in developing economies. Participation in international markets allows achieving economies of scale, mainly due to increased demand. This idea is consistent with Van Biesebroeck (2005) who finds that because of financial constraints and contract enforcement, domestically-oriented firms do not fully exploit scale economies.

7 Wagner (2007a) reviews 54 empirical studies covering 34 countries that address this issue.

8 However, as Eibl and Malik (2016) show, non-tariff measures remain important in sectors with substantial number of politically connected firms.

9 Since its establishment in 1983, the Export Development Bank of Egypt (EDBE), provides short- and medium-term loans to finance capital assets of exporting firms. It also provides bank guarantees required for financing exports. The Export Credit and Guarantee Company, established in 1992 by the EDBE, supports exporters by issuing export credit insurance guarantees covering up to 80% of any losses incurred.

10 These agencies mainly include the Commercial Representation Body, the General Organization for International Exhibitions and Fairs, the International Trade Point under UNCTAD’s initiative, and the Egyptian Export Promotion Centre.

11 In a similar approach Olley and Pakes (1996) use investment to control for this simultaneity issue.

12 This period was obtained by exploiting the information provided for the year preceding each survey year (2004, 2006, and 2008). The World Bank’s Enterprise Survey database is available at: http://www.enterprisesurveys.org/.

13 This feature is consistent with the structure of the Egyptian export sector depicted in the 2005 Trade Policy Review (WTO, 2005).

14 The precision ‘directly’ is important because in developing countries some exporters used to export through a distributor and therefore are much less concerned with the issue of sunk costs. As mentioned above, the sunk costs are at the core of the self-selection hypothesis. In the survey data we will be using, the question on the export status is expressed as follows: What percent of your establishment’s sales in 2007 were (i) sold domestically, (ii) exported directly, (iii) exporter indirectly through a distributer. The responses indicate that 18% of total exporters were indirect exporters.

15 This definition of export-starters, export-continuers and export-quitters, constrained by the limited time period, is more restrictive than that of Wagner (2002), who uses 3-year windows.

16 See Wagner (2002) who estimates the causal effects of exports on firm size and labor productivity in Germany using this matching approach.

17 The same picture is observed in Tables A1 and A2, when the export starter and export continuer dummy variables are used as the treatment variables (see the appendix).

18 Unlike of the Probit model, the estimation of the coefficients of the regressor variables using a Logit model is not affected by the unequal sampling rates for exporters and non-exporters (Maddala, 1992). This problem is frequent in large firm-level datasets, particularly when the number of observation in one groups if is much larger than the number of in the other group. In our case, as it is mostly the case in developing economies, the number of exporters appears to be much smaller than the number of non-exporters.

19 Results are not presented but available upon request.