country 1 (i.e., we vary its multilateral, or overall, degree of openness). Assuming symmet-
ric bilateral trade costs between all countries (τjk = τkj for all j, k) and treating countries
2 and 3 as identical (τj1 = τ1j = τ1 for all j = 1), we solve the model for different degrees
of centrality of country 1 (i.e., we vary τ1) while keeping trade costs between countries
2 and 3 (τjk for all j = 1, k = 1) constant. The outcome is summarized in the following
result and visualized in Figure 3.
Result 2a [Geography and spill-overs]
As the degree of centrality of country 1 goes up, a given rise in country 1’s unemployment
benefits yields a smaller unemployment increase in country 1 and a larger increase in
countries 2 and 3.
For illustration see the variation of τ1 on the x-axis in Figure 3. The variation of
country size on the y-axis is discussed in Result 2b. The figure shows the absolute
change of the unemployment rates, ∆u1 (left-hand diagram) and ∆u2 = ∆u3 (right-hand
diagram) generated by a given change of b1 (from 0.4 to 0.8; hence ∆b1 = 0.4) for different
values of τ1, where lower values of τ1 indicate higher centrality of country 1. For τ1 = 30%
and si = 0.33, an increase of b1 from 0.4 to 0.8 moves u1 up by about 15 percentage points
in country 1 and by about 0.03 percentage points in countries 2 and 3. This effect can
also be read off Figure 2 by comparing unemployment rates at τ = 30% for b1 = 0.4 and
b1 = 0.8.
In line with Results 1b and 1c, the change in unemployment is positive for all coun-
tries. The new insight from Figure 3 is that a higher degree of centrality of the “bad”
country (i.e., a lower τ1) weakens the increase in the unemployment rate in country 1 but
strengthens the increase in countries 2 and 3. We see that the increase in b1 increases
unemployment in country 1 by about 13 percent when τ1 = 0 and by about 16 percentage
points when τ1 = 0.6. Hence, the more central a country is, the lower are the unemploy-
ment costs of its own bad institutions. Trade partners, however, suffer more as a decrease
in τ 1 drives up the change in the unemployment rate. However, quantitatively, the effect
is fairly small.
The intuition for Result 2a is straightforward. If country 1 is more central, it trades
more with countries 2 and 3. If country 1 has no access to international markets (τ1 → ∞),
lower domestic demand for country 1’s products due to higher unemployment in that
country would be tantamount to lower total demand, so that the adverse labor market
implications are most severe. In the other extreme where τ 1 = 0, domestic demand
only accounts for a fraction of total demand faced by country 1’s firms. Since foreign
institutions remain unchanged, there is no first-order effect on foreign unemployment and
hence income, so that the relative downfall of total demand faced by country 1’s firms is
less than proportional to the fall in domestic demand. Therefore, the resulting increase
in the unemployment rate is smaller. That logic holds in reverse for countries 2 and 3
which rely more on country 1’s demand when τ1 is lower.
Next, we study how the size of the “bad” country affects spill-overs. We measure
country size in terms of population, as income is endogenous in our model. More precisely,
we fix the world population Lw = ∑3=1 Li, and then change country 1’s share of world
population from 10% to 90%. The remaining population is distributed equally between
22