mapping from preferences to actions is not unambiguous” (Scheve and Slaughter 2001, p. 4).5
Of course, using survey data has its own limitations. If the Heckscher-Ohlin trade model
accurately describes the world, then that has profound (and negative) implications for the impact of
North-South trade on Northern wage inequality, and equally profound, but positive, implications for
the impact of that same trade on Southern wage inequality. It might well have political implications
for the ease with which the transition economies of Eastern Europe can be integrated into the
European Union. Finally, it would clearly have intellectual implications for the way in which
theorists should specify their models of endogenous tariff formation. Leaving aside measurement
problems in attitude research (to which we return below), survey data on their own cannot tell us
whether the Heckscher-Ohlin model in fact describes reality; all that they can tell us is that agents’
preferences are consistent, or inconsistent, with the predictions of the model. Our claim is that
findings of the latter, more cautious, variety are useful, since the determinants of preferences matter
in themselves, both intellectually and politically.
The plan of the paper is as follows. Section 2 provides a brief discussion of what theory has
to say about the impact of trade on skilled and unskilled wages, since it is this link that leads us to
expect a relationship between skill and trade policy preferences. Section 3 introduces the data,
including our measures of nationalist attitudes, and discusses the strengths and weaknesses of our
evidence relative to the data used by previous authors. Section 4 provides an ordered probit analysis
of the determinants of individual trade preferences in 24 countries, including both ‘economic’ and
‘ideological’ determinants. Finally, Section 5 concludes by drawing some preliminary inferences and
outlining an agenda for future research.
5 Since beginning this project, we have become aware of the simultaneous and independent work
being pursued by Anna Maria Mayda and Dani Rodrik (2001), who use the same ISSP survey data as
we do, and who reach many of the same conclusions.