while a symmetric equilibrium is still stable for the unskilled workers
( {∂ωu / ∂Ui, ∂ωu / ∂Si} should be strictly negative).
In Figures 1 and 216 we plot respectively the low and high-skilled workers real wages
in region 117 at different levels of trade costs. This allows us to analyse how the integration
process, measured by the level of trade costs, will affect the equilibrium size of the
manufacturing sector and the average level of human capital in the regions. We report the real
wage differential for low and high-skilled workers as a function of the share of only one type
of manufacturing labour keeping the other at the symmetric equilibrium. In this way we
assess, for each type of worker, how the real wage differential reacts as a consequence of
variations of both types of manufacturing workers population. Specifically, diagrams (a) [(b)]
of both figures report the real wage differentials as a function of the share of low-skilled
[high-skilled] workers in region 1, given that high-skilled [low-skilled] workers are equally
distributed between the two regions, S = 1/2 [ U = 1/2].
For high trade costs ( φ = 0.05, or τ = 2.714) the symmetric equilibrium is stable since
the cost of supplying a market by exporting is too large. Both low and high-skilled have no
incentive to migrate in region 2. When a worker migrates (either high or low-skilled), the
host region becomes less attractive than the destination one. In the long-run the economy
converges to a symmetric equilibrium in which manufacturing is equally divided.
As the economy becomes slightly more integrated (φ = 0.05, or τ = 2.487), the
symmetric equilibrium is no longer stable for both type of workers. Consider the low-skilled
first. From Fig. 1 it is evident that an increase in the share of both low and high-skilled
16 The model cannot be solved analytically. In the paper, all the figures presented are derived by numerical
simulations. The values of the parameters are similar to those used in related papers. We let μ = 0.3, while σ =
4; mark-up estimates are normally between 20-30%, which correspond an elasticity of substitution σ between 6
and 4. Agglomerative externalities are chosen on the basis of the empirical evidence mentioned in the
conclusions, δ = 0.1. Analytical expressions for φu, φs break and φ self-selection, have been derived using a
procedure first introduced by Puga (1999) to which we refer. Maple files containing the simulation procedure
and the stability analysis are available from the author on request.
24