Appendix 1. Data Appendix
This section provides more detailed description of the data sources used for the analysis in the paper:
Canada’s non-energy exports—The volume data for individual non-energy export products are obtained from national accounts statistics on a quarterly basis (s.a.a.r., chained, 2007 C$). The export price indices are calculated by dividing the nominal export value series by the volume series.
Foreign demand components—Real private consumption, business investment, and government gross investment (federal, state, local) series for the United States all come from the U.S. national accounts statistics (s.a.a.r., chained, 2009 US$). Real domestic demand series, defined as the sum of consumption, gross fixed investment, and government expenditure, for the rest of the world is calculated as the product-specific export share-weighted average of Canada’s 22 export destinations (Australia, Austria, Belgium, Finland, France, Germany, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Korea, Taiwan Province of China, Republic of China, Philippines, Singapore, Mexico, United Kingdom). The original export shares for individual products are from the U.N. Comtrade Database (SITC rev.2, available on an annual basis), which we then re-classify to match the North American Product Classification System and linearly interpolate to obtain quarterly estimates. In the case of the Republic of China, real GDP is used instead of real domestic demand due to data availability, with historical quarterly GDP estimates taken from Abeysinghe and Rajaguru (2004).
Foreign product prices—Foreign product price indices for individual products are calculated as the product-specific export-share weighted PPIs of the United States, euro area, Japan, United Kingdom, and the Republic of China. Due to data availability, we use product-level PPIs for the U.S., euro area, Japan, and aggregate PPIs for the United Kingdom and the Republic of China.
Medas, P. (2012), “Canada’s Loss of External Competitiveness: The Role of Commodity Prices and the Emergence of China,” IMF Country Report, No.13/40, Selected Issues.
Binette, A., D. de Munnik, and E. Gouin-Bonenfant (2014), “Canadian Non-energy Exports: Past Performance and Future Prospects,” Bank of Canada Discussion Paper, 2014-1.
Kejriwal, M. and P. Perron, 2010, “Testing for Multiple Structural Changes in Cointegrated Regression Models,” Journal of Business & Economic Statistics, Vol. 28 (4), pp. 503–522.
Bai, J. and P. Perron (1998), “Estimating and Testing Linear Models with Multiple Structural Changes,” Econometrica 66, 47–78.
Bai, J. and P. Perron (2003), “Computation and Analysis of Multiple Structural Change Models,” Journal of Applied Econometrics 18, 1–22.
Prepared by Minsuk Kim (SPR).
In the rest of the paper, non-energy exports refer to non-energy exports of goods only, and exclude exports of services.
A recent IMF study estimates that a 10 percent increase in the REER results in about 1½ percentage point decline in Canada’s U.S. market share based on the sample period of 1975–2010 (Medas, P., “Canada’s Loss of External Competitiveness: The Role of Commodity Prices and The Emergence of China,” 2012 Article IV Consultation Staff Report, Selected Issues Paper).
Final domestic demand is defined as the sum of consumption, gross fixed investment, and government expenditure. For China, we use the gross domestic product instead due to the lack of available data. We exclude the foreign demand term for “passenger cars and light trucks”, “medium and heavy trucks, buses, and other motor vehicles,” and “aircraft, aircraft engines and part” as the U.S. share accounts for very close to 100 percent.
We use OLS estimates throughout this paper, which have non-standard limiting distributions when the series are cointegrated I(1)s as in this case. However, other more suitable estimation methods (including Dynamic OLS, fully-modified OLS, VECM) also produced point estimates very close to those of the OLS. It should be noted, however, that the OLS appears to underestimate the size of the coefficients for the real effective exchange rate and the U.S. consumption relative to these other methods.
The strategy involves first estimating l break points
The list of winners and losers identified in our analysis overlaps on many accounts with a recent study by Binette, de Munnik, and Gouin-Bonenfant (2014) from the Bank of Canada, but there also exist some notable differences (e.g. pharmaceutical). This could reflect among others the different time horizons considered and the choice of benchmarks.
Given the parsimonious specification used in this paper, we cannot rule out the possibility that the identified structural breaks were in fact due to the influences of other potentially important but omitted variables (e.g. those capturing non-price export competitiveness). To check the robustness of the timing of breaks against this possibility, we have run the same break test on our baseline equation augmented by the polynomial terms of the estimated Y’s as proposed in the Ramsey RESET test. Even with the additional terms, however, the Bai and Perron test still indicates a structural break in 2007 for total non-energy exports, providing an indication that the break point is robust to the potential omitted variable problem.