Therefore, an important question is: Are they differences between wages outcomes of migrants and
non migrants? It there are, it is possible to explain them on the basis of search processes with a spatial
dimension? We deal with this question in section 3, comparing wage equations between migrants and
non migrants. However, in econometric analysis, we must recognise that a migrant differs from an
immobile entrant not only because he has found better opportunities elsewhere than locally. He can
also differ because he does not face the same opportunities, at his origin place and abroad. Therefore,
migrants may be self-selected, an issue that can be analysed using the Roy model (Borjas et al, 1991).
From an econometric point of view this leads to the two-step Heckman procedure (Heckman, 1979).
The results of estimation are presented in section 4. After a brief comment on the decision to migrate,
we insist on the impact of migration on wage. We show two complementary selection effects: a
positive selection effect on migrants and a negative one on non-migrants.
Section 2. Job search processes in space.
In the typical job search model (Lippman and McCall, 1976), an agent looks for opportunities
characterised by the expected value of the wages flow over all future periods, v. At period t, the agent
entails a search cost equal to ct and draws a random offer in the cumulative distribution Ft ( v). His
optimal strategy is to accept the offer if v exceeds a reservation value Vt+ 1, equal to the value of
continuing search at least at period t + 1. The series of reservation values is recursively determined
using the Bellman equation,
+∞
v = - ct+ ∫ max(v, Vt+1 dF (v ) (1)
0
A particular case of (1) is when the lifetime horizon is infinite and all periods are identical. Then,
ct = c , F1 = F, and Vt = V = β- w , where β is the discount factor and w is the reservation wage,
id est the minimal accepted wage. The distribution of accepted wages is
G(w) = [1 - F(w )] 1 [F(w) - F(w)]. All changes implying a higher return from future search, for
example lower search costs, also imply a higher reservation wage and a higher mean accepted wage.
The model may be easily generalised to a spatial setting, with searchers moving between local labour
markets. When migration is speculative, the agent move before search. Each local labour market, n, is
characterised by the local search cost cn and the local distribution of wage offers, Ftn (v), hence by
the local value of search, Vtn . The agent chooses to locate where the net returns from search,
Vtn - mn (where mn is the migration cost to n), is highest. On the contrary, the agent may stay on at
his origin place until he has received an acceptable opportunity. The global search cost is the sum of all
search costs on each local market, ct = ∑n cn . The global distribution of wage offers combines all
paper.