that have to be faced: not all DCs are homogenous. Instead, they differ greatly in their endowments of
capital, labor and skills. Davis (1996) points out that in a 2-factor (capital and labor) world, a country
may be globally labor-abundant, but locally capital-abundant, in the sense that it is capital-abundant
relative to other countries in the same “production cone”; for such a country, liberalization lowers
wages. By implication, a middle-income country such as Mexico might be skill-abundant relative to
countries like China and India; they might therefore protect their unskilled-labor-abundant sectors;
and they might thus see skill premia rising on liberalization (Slaughter 2000; Wood 1997). There is in
fact evidence that unskilled-labor intensive sectors received the most protection in countries such as
Mexico and Morocco priorto liberalization (Currie and Harrison 1997; Hanson and Harrison 1999).
The argument may have some relevance for this study. With the exception of the Philippines,
our data set does not include any third world countries (see below); rather, it contains data for rich
countries, and forthe transition economies ofEastern Europe. We will be interested to see whether
there are different determinants of trade policy preferences in the latter group of economies than in
the former; but the Davis article reminds us that while these countries may be poor relative to the
West, they are rich, and skill-abundant, relative to most of the rest of the world. In that sense, our
sample is a truncated one, and the results need to be interpreted in that light.
In particular, Heckscher-Ohlin theory predicts that the highly-skilled will favor free trade in
the most skill-abundant countries (such as the USA): in these countries, a regression explaining
protectionist attitudes should find a negative coefficient on skills. Whether the coefficient on skills in
the least skill-abundant country in our sample should be positive or negative depends entirely on
where that country fits in terms of the world-wide hierarchy of skill-abundance; the sign of the
coefficient is thus apriori unclear. In order to test Heckscher-Ohlin theory, therefore, we will pursue
two strategies. First, we will estimate models (for the entire sample of countries) of the form:
PROTECTij = αi + β1SKILLj + β2SKILLj*GDPCAPi + β3Xij + εij (1)