Notes

^1.

See Macdonald (2007a, 2007b).

^2.

See Macdonald (2010).

^3.

In the literature surrounding the System of National Accounts (SNA), the trading gain is derived by deflating net exports directly rather than using an implicit price deflator. The SNA presents several options for deflating net exports, including import prices, export prices, an average of import and export prices or a final domestic expenditure price index. Discussions of these alternative methods can be found in Geary (1961), Stuvel (1956), Denison (1981), Nicholson (1960), Courbis (1969), Kurabayashi (1971), Silver and Mahadavy (1989), the SNA 1993, Kohli (2004), and Macdonald (2007b).

^4.

This distinction may not be quite as clear-cut as it appears, because while the OECD primarily makes reference to a production concept, it suggests that it can be used to examine welfare—though the latter is qualified as meaning GDP per head and therefore probably just refers to production capacity.

^5.

See the ICP 2003-2006 Handbook, chapter 1, p.1.

^6.

See World Bank, International Comparisons Project 2003-2006 Handbook, Technical Notes.

^7.

Eurostat produces annual values of PPPs.

^8.

Commodity bundles may differ because of differences in consumer taste and differences in production systems. Matching investment commodities is particularly problematic when it comes to buildings and engineering projects.

^9.

This route is less direct than the UV approach since the expenditure programs that yield the average price data that are mapped to the industry level come mainly from consumer surveys, while the commodity categories from the industry accounts come from output data derived from firm-based surveys. Definitions of commodities are not always the same in each survey.

^10.

Except in special circumstances where the law of one price holds for both imports and exports.

^11.

One way to evaluate the size of the bounds that should be used is to compare two estimates that use the same data, but are produced by two independent teams. Baldwin, Gu, and Yan report that Rao, Tang and Wang (2004) make use of the same commodity data from the Canadian input/output tables and slightly different assumptions about commodity splits by industry and obtain a PPP estimate of 0.88 compared to the 0.85 estimate produced by Baldwin, Gu, and Yan used here.

^12.

For a recent academic article that recognizes the differences, see Feenstra et al., 2009.