A second key point about Deardorffs six SS versions is that the best guide for empirical work depends on the world’s dimensionality. The first three versions hold only in worlds with two factors and two products. The Strong Version holds only for worlds with the same number of factors and products (more than two allowed) and only under certain technology restrictions. Only the last two versions hold for any arbitrary number of factors and products. If one thinks that the world cannot be reasonably approximated as 2×2 then one should focus on the latter three versions. If one is uncomfortable assuming the world is even then one should focus on the last two versions.
The third key point is that only the Correlation Version directly addresses something other than real-wage changes. This matters because all nine product-price studies surveyed below focus on rising inequality, not the absolute losses of the less-skilled.3 In light of the discussion regarding dimensionality, this point suggests that the Correlation Version might be the best guide for empirical work on relative wages. Friends and Enemies focuses on the price change of only a single product, and it claims only that that single price change will raise the real return to one unidentifiable factor an lower the real return of some other unidentifiable factor. In contrast, the Correlation Version relates any vector of relative product-price changes and allows these changes to be systematically related to relative factor-price changes to relative factor prices. Granted, one cannot predict for certain what will happen to any particular relative factor price. But one can say that “on average,” factors employed intensively in rising- (falling-) price industries will experience relative price increases (declines). This result seems closest in spirit to the empirical issue of rising wage inequality. loans