states opt to enter a PTA. One way to account for any unmeasured heterogeneity of this sort is to
include country-specific or pair-specific fixed effects in the model. Various observers argue that
in situations like ours, where the dependent variable is dichotomous and has highly skewed
distribution, a fixed-effects specification introduces more problems than it addresses (for
example, Beck and Katz 2001). However, these observers maintain that if researchers are
committed to estimating a fixed effects model when analyzing such a dependent variable, it is
preferable to include country-specific rather than pair-specific effects. To assess the robustness
of our results, we therefore re-estimate equation (1) after including country-specific fixed effects.
The results presented in the final column of Table 3 show that the estimates of Veto
Points
i
and Veto Points
j
continue to be negative and statistically significant. Equally, these
estimates are only marginally smaller (in absolute value) when a fixed-effects specification is
used. Taken together, the results in this section therefore demonstrate that our earlier findings
are quite robust.
Conclusions
The recent proliferation of PTAs has stimulated rising interest in the consequences of
these arrangements. Far less effort has been made to assess the factors prompting states to enter
PTAs, especially factors operating in the domestic political arena. In this paper, we have
addressed this issue by analyzing the role played by domestic veto players. Our model suggests
that it becomes increasing difficult for states to forge PTAs as the number of veto players rises.
A series of empirical tests, based on an analysis of PTA membership from 1950-1999, support
our argument. Countries are much less likely to sign trade agreements as the number of veto
players increases.
28
Convention