ABSTRACT
This note revisits the role of migrant social networks (MSNs) as determinants of bilateral migration flows. We do so using two alternative databases, covering about 190 world countries for the period 1960–2000. In line with previous estimates, we consistently find a significant, positive, impact of MSNs at destination. We also find that such an impact decreases very slowly as the year when MSNs are observed is farther back in the past. We also find that geographical distance is becoming less and less important over time but significantly explains the absence of a migration corridor only when MSNs at destination have very small sizes.
Acknowledgements
Thanks to Marina Mastrorillo and Gianluca Orefice for useful suggestions. All usual disclaimers apply.
Notes
1 Migration flows are estimated via a statistical procedure aimed at predicting the number of movements required to meet the changes over time in migrant stock data, using an iterative proportional fitting algorithm (Deming and Stephan Citation1940).
2 The slope of a pooled regression between stock-difference and Abel flows is 0.34 (with a t-test of 59) and an .
3 In line with Beine and Parsons (Citation2015), we do not find any consistent and significant effect of climatic anomalies on migration flows, when they are inserted in the migration flow equation. Instead, they are a valid and strong instrument for both per-capita income and migration stocks.
4 We have also tried to use additional covariates (e.g. unemployment, relative quality of institutions, Marshall and Jaggers Citation2010, etc.), but they either substantially reduced our sample size or turned out to be consistently not significant.
5 This result is robust to inclusion of higher-order income terms (Mayda Citation2010). We did not account for poverty indicators as in Belot and Hatton (Citation2012) because that would have substantially reduced the sample size.
6 The probability of a zero flow given a positive stock also increases the smaller population and per-capita GDP at origin, if the origin did not experience violence episodes, and if the two countries do not share a common language. Note also that the probability of a strictly positive flow in absence of network effects (S = 0) decreases, as expected, the farther away the two countries are.