Do the Return Intentions of French Migrants Affect Their Transfer Behaviour?

Abstract

This paper investigates to what extent the return behaviour of migrants affects their transfer decisions, both at the extensive and intensive margins. We use a unique data set collected on migrants aged 45–70 living in France, with detailed information on both return intentions at retirement and on private transfers. We find that the temporary nature of migration strongly influences the pattern of transfers made by migrants. The probability of remitting for either personal savings or to family members in the origin country increases by more than 10 percentage points with return plans, the latter having no effect on gifts to family members living in France. At the intensive margin, the amount of personal savings sent to the origin country is about twice as high for migrants who intend to return.

Acknowledgements

We would like to thank the editor, two anonymous reviewers and seminar participants in Angers for their very helpful comments and suggestions on a previous draft. Data and Stata code are available upon request from the author.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Global remittance flows, including both transfers to developing and high-income countries, were estimated to be $526 billion in 2012. By decreasing order of importance, the top five recipient countries of migrant remittances were India, China, the Philippines, Mexico and Nigeria.

2. For some outcomes, it may be important to differentiate between migration and remittances. For instance, the receipt of remittances is expected to increase human capital investment in young children, while migration may have harmful effects due to parental absence. For a discussion and empirical evidence, see McKenzie and Rapoport (2011).

3. Dustmann and Mestres (2011) analyse savings and asset holdings of immigrants in relation to their return plans. They do not find that immigrants with temporary migration plans save more than immigrants with permanent migration plans. Dustmann and Weiss (2007) examine the different motives for a temporary migration.

4. This survey was conducted by the Caisse Nationale d’Assurance Vieillesse with the help of the Institut National d’Economie et de Statistique. It is representative of the older migrant population living in France in 2003. The lower limit of 45 for migrant age was chosen because the focus of this survey was on retirement (Attias-Donfut, Davaut, Gallou, Rozenkier, & Wolff, 2006).

5. We choose to combine the decisions of both making return trips and returning permanently, as they reflect a strong attachment to the origin country.

6. van Dalen and Henkens (2013) show that health matters when explaining differences between the behaviour of movers and of respondents who stated intentions to emigrate but did not move during the period under consideration.

7. At most, four transfers per respondent were recorded in the PRI survey. For an overview of the pattern of private transfers from migrants living in France, see Wolff, Spilerman, and Attias-Donfut (2007).

8. 6.5 per cent of migrants have no child; while 9.5 per cent have children both in France and in the origin country.

9. On average, the amount sent is substantially higher for savings (5970 euros per transfer) than for remittances (1785 euros) and family transfers (3320 euros).

10. The difference in transfer values between those who intend to return and those who intend to stay in France is 190.80 euros for remittances to origin country. The gap is −912.6 euros for family transfers in France, but the null hypothesis of equal means cannot be rejected at the conventional level (the critical probability is 0.147).

11. These additional results are available upon request. The statistics associated with the Wald tests are, respectively, 1.27 for personal savings to origin country (with a probability of 0.261), 2.02 for remittances (with a probability of 0.156) and 0.02 for family transfers in France (with a probability of 0.886).

12. Since the transfer variable is expressed as a logarithm, the marginal effect associated with the estimated return coefficient ${\hat{\delta }}_k$ is given by $exp\left({\hat{\delta }}_k\right)-1$.

13. Location choices will even be totally uncorrelated with the dependent variable if they contain only random errors.

14. In a previous version, we used return decisions of immigrants from the same country of origin as an instrumental variable to identify the effect of individual return intentions on remittances. As pointed out by a reviewer, both return intentions and remittances of immigrants can be directly affected by return decisions of immigrants from the same country of origin, and there is no mechanism (such as randomisation) which ensures that return intentions of immigrants are the only channel through which return decisions of other immigrants affect remittances.

15. Also, we cannot disentangle how past transfer decisions or past return intentions affect current behaviours of migrants.