Foreign education and international trade: empirical evidence from selected Latin American countries

Abstract

Social ties among university students – of friendship, mutual trust and attachment to the alma mater – tend to be robust and enduring. Through information-diffusion and behaviour-enforcement mechanisms, they can boost the economic exchanges between countries. This paper tests the influence of Latin American people with a tertiary education in OECD countries on the bilateral trade between the home economy and the country of the alma mater, taking into account potential endogeneity concerns. Results show that Latin American student networks exert strong, positive and significant effects on bilateral imports and exports. A 10% increase in the number of Latin American students in the OECD economy boost bilateral trade by about 3%. At a more disaggregated level, their impact on differentiated goods is significantly higher than on homogenous products. Their incidence is lower in the presence of bilateral trade agreements and economic integration between countries. Results are robust to the deep economic and political transformations of the period considered, and to the use of different regressors and specifications.

Keywords:

• Bilateral trade
• education networks
• international students
• Latin America

JEL Classification:

• F14
• F29
• F59
• I20

Acknowledgements

I thank Maria Luisa Recalde and Pedro Degiovanni for their participation in an earlier stage of this research. I also thank for their comments Laura Marquez-Ramos, Sophia Dimelis, Pedro Moncarz, Benedikt Heid, and the participants at the International Trade Study Group 2015 and the Arnoldshain Seminar 2015.

Notes

1. The empirical data show that the great majority of international students return home after completing their studies. In 2006–2008, about 75% of all international students in selected OECD countries returned home after completing their studies (OECD 2011).

2. Using data on migration, Docquier and Rapoport (2012) find that Caribbean and Central American countries are strong exporters of human capital. As suggested by one referee, this may imply that students from these countries educated in the OECD destinations can be expected to strongly influence bilateral trade with the home economies.

3. The partial shift in student destinations may be related to several factors. First, after the end of the cold war, the United States had less political incentives for attracting students from Latin America; second, the competition of South East Asian students for education in American universities increased; third, entry policies in the United States after the terrorist attacks of September 2001 became more restrictive; finally, the costs of university fees in the United States grew more than in Europe, Australia and Canada (OECD, 2011).

4. During the last two decades, the lower growth rates of trade between the Latin American and OECD economies is partly compensated by more trade between Latin American countries and South-East Asia (Dosch and Jacob 2010).

5. Standard errors in all regressions are heteroskedasticity robust and clustered by country-pair. Zeros in the dependent variables are 0.06% regarding exports and 0% for imports; however, following Silva and Tenreyro (2006) and Head and Mayer (2014), the PPML-FE estimator is used as a further test of the gravity equation. In columns 9 and 10, I use the two-step Sys-GMM specification in levels and differences, which accounts for multilateral resistance factors (Martìnez-Zarzoso and Nowak-Lehmann 2009). Moreover, time-invariant dyadic variables, such as Distance and Language, are included among regressors (Roodman 2009).

6. Potential endogeneity is detected by the Durbin–Wu–Hausman test: the error terms of the base regressions of columns 1 and 2 (Table 1) and International students are significantly correlated. On the other hand, test results (Arellano Bond and p-values of Hansen tests) reported in Table 1 attest the validity of the Sys-GMM results (columns 7 and 8).

7. The level of democracy in the average Latin American country rises from 0.34 to 7.46 (it increases in all selected countries except Venezuela, where it falls), Civil liberties rise from 4.35 to 5.18 and Polity rights by about 30% (the latter in Table A1).

8. The fixed-effect estimator is more efficient than the first difference estimator if the error terms are serially uncorrelated. However, the first difference estimator should be used if only the first differences in the errors are uncorrelated (Wooldridge 2002).