How Robust Are “The Facts”? Methodological Issues
The variety of empirical methods used in this research area raises the issue of how robust results are to various methodological choices. In this section I discuss how sensitive results are to four different robustness issues: the selection and weighting of industries sampled; the decade considered; the extent of data aggregation; and the measurement of skills.
Robustness to Industry Selection and Weighting
The SS theorem’s zero-profit conditions apply to all domestic industries currently producing positive output. In the matrix equations (1) and (2) there are N rows corresponding to all N industries currently being produced. Any domestic industry not currently producing is assumed to have average costs exceeding industry’s price; thus that industry has no zero-profit equality. This suggests that empirical work on the SS theorem requires data on all operating domestic industries.
This raises two separate issues. One is whether to limit the sample to just all operating domestic industries that are tradable. As discussed in Section 2, four of the versions of the SS theorem do not explicitly involve international trade. The link between product prices and factor prices holds even in autarky. Given this, focusing on just tradable industries seems appropriate when trying to understand product-price changes attributable to international trade.7 The other issue is conditional on a selected sample, are all available industries included in the data analysis. Theory suggests that all data should be included. Missing industries introduces the risk of not being representative of the appropriate full sample.