Abstract
FOLLOWING the seminal work of Nelson (1959), recent studies of migration have emphasized the role of prior migrants in influencing the settlement patterns of subsequent migrants. Nelson hypothesized that the new migrant is more likely to receive information about those regions populated by relatives and friends,' and that he is more apt to receive temporary accommodations and to feel more at home in those areas. Therefore, it is anticipated that new migrants will be attracted to those destinations inhabited by earlier migrants from the same region of origin. Greenwood (1969, 1970), Vedder and Gallaway (1972), and Levy and Wadycki (1973) among others have attempted to account for these and effects by including a migrant stock variable in their explanatory models. In general, these authors have obtained positive and highly statistically significant coefficients on the migrant stock variable. The proper interpretation of these results, and of the migrant stock variable, however, is not a settled matter. Migrant stock may emerge as a regressor from either of two models of the migration process. It may represent the influence of family and friends or it may capture a partial adjustment mechanism. Further, it may introduce multicollinearity into the statistical results. The main elements of these issues were raised by Laber (1972) in his comment on the Greenwood papers. Laber argued not only that Greenwood inadvertently estimated a lagged adjustment but also that on the basis of his (Laber's) model Greenwood's estimates obtained incorrect algebraic signs. At issue then is whether the alternative models are representative of the migration process. The uncertainty regarding the role of migrant stock is largely due to its correlation with other explanatory variables. Broadly, migrant stock is the sum of all past gross migration less deaths and secondary migration of the earlier migrants. Migrant stock, therefore, is itself a function of all those factors which influenced the earlier migration. As a result it may act as a proxy for lagged explanatory variables. Further, to the extent that current values of the other independent variables are correlated with their lagged values, migrant stock will also be correlated with these current explanatory variables, possibly resulting in serious multicollinearity. Omission of the migrant stock variable, on the other hand, may result in specification bias as well as in the loss of information regarding the family-friends effect. Discussion of the interpretation and role of migrant stock has been primarily in terms of the consistency of Greenwood's model, and in terms of the statistical problems of specification bias and multicollinearity. As a result, the original interest in the family-friends effect was soon obscured. Greenwood, for instance, after briefly discussing the rationale for using migrant stock, devotes much of his theoretical evaluation to econometric questions.2 Following Nelson, Greenwood demonstrates that omission of the migrant stock variable in empirical work will, because of its relationship to the determinants of prior migration, cause the parameter estimates to overstate the true effect of the variables. Thus, interpretation of Received for publication September 26, 1975. Revision accepted for publication May 10, 1976. * The authors thank Raouf S. Hanna and Walter J. Wadycki for the helpful comments and suggestions they have made on earlier drafts of this paper. An earlier version of this paper was presented at the Annual Western Economic Association Conference, June 26, 1975. Responsibility for remaining errors and omissions, of course, rests with the authors. 'The information could be of a discouraging nature rather than encouraging. However, the current presence of the prior migrant in the destination region seems to increase the likelihood of the information being positive. 2This is true of both the 1969 and 1970 papers.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have