Assume therefore that τ falls below 0.5. The first difference from Section 2 is that this
reduces the absolute profitability of exporting and not just its profitability relative to foreign
direct investment. The total profits from exporting are now:
∏x = nπ[f,(n-l)τ] (26)
Setting this equal to zero, the threshold external tariff t is increasing in τ, and hence it falls
as τ falls. In words, reducing internal barriers makes exporting less attractive for a new
reason, additional to those already noted in Section 2. Competition from partner-country
firms in each union country is increased, which makes it less profitable for the multinational
to supply any union countries externally. With linear demands, the threshold external tariff
is:
( = l+(n-l)τ (27)
n+1
This falls from a maximum of 0.5 when internal and external barriers are equal, to a
minimum of 1/(n+1) when internal barriers are abolished. In Figure 4, the vertical locus
separating the X and O regions shifts to the left.10
Turn next to the FDI case. As in the previous sub-section, we need to distinguish
between two cases. If τ is below ⅛, then the multinational faces competition from all other
union firms in every market where it locates. By contrast, if τ exceeds ½ (while still lying
below ½), then firms from other union countries are not competitive in any market where the
multinational locates. In this case, FDI leads to the multinational behaving like a duopolist,
10 A complication arises with non-linear demands. It is shown in the Appendix that the
derivative of π with respect to t could be positive if demand is highly convex. This in turn
raises a further possibility that the derivative could change sign as falls in t alter the curvature
of the demand function. As a result, the X and O regions could overlap in Figure 4. This
possibility must be considered rather esoteric, and I ignore it henceforward.
15