Abstract
We provide regressions for the net immigration flows of developing countries. We show that (i) savings finance emigration and worker remittances serve to make staying rather than migrating possible; (ii) lagged dependent migration flows have a negative sign in the presence of migration stock variables; (iii) stocks of migrants in six OECD countries and in the developing countries have non-linear effects. Some of the non-linear effects of the economic variables vanish if indicators for disasters, conflicts and political instability are taken into account but new ones come in for these latter variables.
Acknowledgements
The author is grateful to an anonymous referee, Bertrand Candelon, Femke Cramer, Huub Meijers, Pierre Mohnen, Eddy Szirmai, Bart Verspagen for useful comments, and participants of a UNU-MERIT seminar and the research day of the Netherlands Network in Economics (NAKE) for interesting discussions.
Notes
1With the better availability of bilateral data this can be improved. But bilateral data are not available, for example for remittances. They are currently constructions transforming balance of payments data of countries into bilateral information by the use of models (see Ratha & Shaw, Citation2007).
2Naudé (Citation2009a,Citationb) uses number of disasters rather than number of people affected and instead of battle deaths he uses number of years of conflicts. We think that the seriousness of the events is taken into account better in the variants of the variables we use.
4 http://www.emdat.be/. Emergency Events Database of Centre for Research on the Epidemiology of Disasters.
5 http://www.systemicpeace.org/inscr/inscr.htm (see Marshall & Jagger, Citation2009). There we also found data on major episodes of political violence (MEPV) from ethnic, civil and international conflicts. The indicators actotal and totalac are aggregates from subcomponents that have been given values from 1 to 10 (see Marshall, Citation2009). They are highly correlated with the data on forcibly displaced persons and with the number of battle deaths and therefore will not appear in the regressions shown. Similarly, we could not find any effect for refugees by country of origin and country of asylum, probably because they are included in the international stock of migrants.
7See also Baltagi (Citation2008, chapter 8).
8By implication, other studies often covering longer periods can do so if they do not employ lagged dependent variables and the even further lagged instruments, which costs two five-year observations and if they do not use the migration stocks in the six OECD countries, which are available only since 1975 (rather than 1960).
9The panel average of the migration stocks as a percentage of the labour force of the country of origin is 2% for the small and 8.7% for the richer and 5% for the large sample. The standard deviation is 7.65%, 14% and 11.8% respectively. The maximum values are 85%, 77% and 85% respectively.
10Hatton and Williamson (Citation2005, Table 2.5) for net immigration, 1970–2000, in a regression for 80 countries, some of which are developed countries, and Clark et al. (Citation2002, Citation2004, Citation2007) for migration into the USA from 81 countries, 1971–1998 also use a migration stock variable. But whereas ours is the stock of migrants in the six OECD countries with origin in a developing country, they use the stock of migrants in the developing country born in a different country, which we will use in regressions 4 and 5. They find a linear, positive and linear-quadratic inverted U-shape, respectively.