Notes

^1.

For an in-depth treatment of head office locations and foreign multinational corporations in Canada, see Baldwin and Brown (2005).

^2.

With the exception of block-faces, the six-character postal code is the most disaggregated geographical unit available (there are more than 750,000 postal code units across the country (Statistics Canada 2002)).

^3.

The $1,200 cutoff ensures that the sample covers only those workers actively engaged in the labour market.

^4.

These synthetic establishments represent the average values for shipments, total employment, value-added, and other key characteristics of the initial 13,000 ASM plants considered for the mapping. As such, they represent combinations of ASM plants found across industry-geography cells. To create these establishments, we use an imputation method based on a series of underlying assumptions relating to the defining characteristics of actual ASM plants. For example, if two plants within the same industry-geography cell are (i) exporters or non-exporters, (ii) foreign-controlled or domestic, (iii) part of multi-unit or single-unit production facilities, or any paired combination thereof, then the synthetic establishment represents their average values. More details relating to this imputation method are available in Breau and Rigby (2010).

^5.

These broadly-defined industrial sectors, along with the use of aggregate regions, are adopted in order to preserve confidentiality when disclosing results for certain units with smaller sample counts.

^6.

The foreign-control status of a plant in the 1999 ASM is based on the Corporations and Labour Unions Returns Act-now the Corporations Returns Act-definition of foreign control. The latter considers a corporation as foreign-controlled when either 'direct' or 'effective' control (i.e., usually when more than 50% of the voting equity of a corporation) is held by a person, group, or corporation not resident in Canada (Statistics Canada 2005).

^7.

The capital-to-labour ratio used in the analysis is proxied by the difference between manufacturing value-added and the wage bill for production workers divided by manufacturing value-added for each establishment. Following Baldwin and Gu (2003), a three-year-mean of the capital-to-labour ratio is also used to eliminate some of the volatility which inevitably accompanies such data. Multi-plant is a binary variable that takes on a value of 1 (one) when an establishment is part of a firm with two or more plants and a value of 0 (zero) otherwise.

^8.

Metropolitan areas are defined as the 47 tracked census metropolitan areas and census agglomerations used in the 2001 Census of Population.

^9.

That is, controlling only for plant size and capital intensity produces point estimates for the export and foreign-control status dummies that are qualitatively unchanged.

^10.

Note also that post-estimation diagnostics of variance inflation factors reveal no issues of multicollinearity across the independent variables for all models.

^11.

Note also that the coefficient estimate on the export binary variable for the Prairie provinces is statistically significant only at the .10 level.

^12.

The logic is as follows: the costs of exporting to non-U.S. markets (e.g., European and Asian markets) are higher than the costs of exporting to the U.S. market. Plants that export to those non-U.S. markets should therefore be more productive than those plants that export to U.S. markets in order to overcome the higher costs of entry into those markets. As more productive plants generally pay higher wages, those plants that export to non-U.S. markets will also pay higher wages. We thank an anonymous reviewer for this point of clarification.