assume that all searchers face the same migrations costs and the same distributions of wage offers
both locally and abroad. Only differences in search costs induce differences in reservation and
accepted wages. Migrants face a lower search cost abroad than at home, hence a positive difference
between values of search abroad and at home, large enough for covering migration costs. The opposite
holds true for non migrants.
When search costs at home, ch , and abroad, ca , are independently distributed among agents, or
negatively correlated, migrants face higher search costs than non migrants at home, and lower search
costs than non migrants abroad. Therefore, were they staying at home, migrants would accept lower
wages at home than non migrants; and, were they moving, non migrants would accept lower wages at
home than migrants. The latter are negatively selected at their origin and positively selected at their
destination and, as in the homogenous case, migrants earn higher wages than non migrants.
This is no longer true with search costs at home and abroad positively correlated. Higher search costs
at home do imply higher search costs abroad. When the higher the search cost at home, ch , the higher
the difference between search costs, ch - ca, migrants have high search costs both at home and
abroad. Therefore, they are negatively selected and accept lower wages than non migrants, both at
home and abroad. Conversely, when the higher the search cost at home, c h , the lower the difference
between search costs, ch - ca, migrants have low search costs both at home and abroad. They are
positively selected and accept higher wages than non migrants, both at home and abroad.
Besides intrinsic differences among individual, the search process itself may be a major source of
heterogeneity and then of self-selection. As search is going on, agents revise their expectations and the
costs they face evolve. After a long duration of unsuccessful search at home, a searcher expects the
distribution of local job offers to worsen. Moreover, financial constraints may tighten, implying higher
search costs. The value of continuing search at home and the reservation wage move downward. It the
searcher was not considering search abroad before, he will accept it in the future. In this context, long
search duration imply negative self selection of migrants.
Therefore, the links between migrants and non migrants wages of previous searchers cannot be
analysed without controlling self selection effects.
Section 3. Measuring the impact of migration on wage
An empirical framework for measuring these effects thus begins with the specification of the wage
equation. In addition to the classical determinants likely to influence wage such as gender, marital
status, educational level and job characteristics, we introduce the notion of migration:
W = Xβ + cCM + u∖
(1)