Tables 5 and 6 explore the links between skill levels and preferences in greater detail.23 The
first regression in Table 5 uses five dummy variables reflecting the five skill categories, rather than
the single high-skill variable used in Table 4, and as before introduces them into the regression both
on their own, and interacted with GDP per capita. Using this finer grid does reveal several
relationships consistent with Heckscher-Ohlin theory. Most importantly, the coefficient on Skilll is
negative, suggesting that the lowest skilled are in favor of free trade; but the interaction term is
positive, suggesting that the association between being low skilled and having protectionist
preferences rises with incomes. The coefficients imply that in countries with per capita incomes
below $11942 in 1995 (roughly $1000 more than incomes in Slovenia), belonging to the lowest skill
category is associated with a preference for free trade, but that in countries with incomes higher than
that amount those in the lowest category are more likely to be protectionist. We take this to be strong
evidence in favor of the Heckscher-Ohlin hypothesis. A similar sign pattern emerges for Skill2,
although the coefficient on Skill2 itself (as opposed to its interaction term) is statistically
insignificant. For Skill3 and Skill4 (but not Skill5), the sign of the interaction term becomes negative,
again consistent with the theory, although the coefficients are statistically insignificant.
The second regression drops the interaction terms, but allows the coefficient on the high-skill
variable to vary, depending on whether the economy is western, a transition economy, or the
Phillippines. In the West, high skills are strongly associated with a preference for free trade. They are
also strongly associated with such a preference in the transition economies, although the coefficient is
only about half the size as in the West. Finally, the relationship in the Phillippines is insignificant.
Again, this is suggestive evidence in favor of the Heckscher-Ohlin world view.
Table 6 pursues the same issue in a different way, again using the single ‘high-skill’ dummy
variable. It runs identical regressions (data permitting) for each country individually. While
interesting patterns emerge for several of the other variables, our focus here is on the skill variable.
Figure 1 plots the ‘high-skill’ coefficient, for each country, againstthat country’s GDP per capita. A
23 Country dummies were included for all regressions in Table 5, and Tables 7 through 9.
17