where PROTECTij is the extent to which individual] in country i is protectionist; SKILLj is a
measure of the individual’s skill; GDPCAPi is the GDP per capita of country i; and Xij is a vector of
control variables.6 In this set-up, the test of whether Heckscher-Ohlin theory holds is the sign of the
interaction term, β2, which should be negative, since in richer countries high-skilled workers should
be more in favor of free trade (i.e. less protectionist).7
Second, we run country-specific regressions of the form
PROTECTj = α + β1SKILLj + β2Xj + εj (2)
and compare the β1 coefficients across countries. Again, the test ofHeckscher-Ohlin theory is
whetherthese coefficients are systematically lower (i.e. more negative) in richer, more skill-abundant
countries.
3. The data
What do we need to accomplish our objectives? We need a data set that provides information
on individuals’ trade preferences, socio-economic position, socio-demographic characteristics and
political attitudes. Since the Heckscher-Ohlin model predicts that skill levels will have different
implications for trade policy preferences in different countries, the data should be cross-national in
scope.
What we have are data provided by the 1995 International Social Survey Programme (ISSP)
module on national identity. The ISSP is an international consortium of survey research agencies that
6 Equations (1) and (2) are used for expositional purposes; however, because of the nature of the data,
we actually estimate non-linear ordered probit models, as explained below. As shown the
specification incorporates country dummies, but we also experiment by running regressions without
these country fixed effects.
7 We used country GDP per capita rather than educational attainment variables for reasons given in
Section 4; this amounts to assuming that GDP per capita is highly correlated with country
endowments ofhuman capital.