Note on Research Methods: BC Carbon Tax 2014
We examined changes in the use of fossil fuels since BC’s carbon tax shift took effect on July 1, 2008, using the latest data from Statistics Canada (through 2013). Annual results are given for the period July 1–June 30, to better isolate the effects of annual increases in the carbon tax rate (which took effect July 1 each year). This is why results are given for hybrid years (e.g. 2008-9). We focus only on those fuels subject to the tax (it exempts a few smaller sources, like asphalt, and most aviation fuel use). We also compare BC with the rest of Canada, which helps to factor out any effects resulting from GDP changes (and other common effects). To factor out population change effects, the comparisons are made on a per capita basis.
The fuel use figures come from Statistics Canada’s dataset Supply and Disposition of Refined Petroleum Products (CANSIM 134-004). Population figures come from CANSIM 051-0001. GDP figures come from CANSIM 379-0030. (Note that GDP results are given by calendar year, rather than July 1-June 30 year, due to constraints in data availability.)
Alert readers may note that some of the figures for prior years may have changed a bit. This is for two main reasons. First, Statistics Canada periodically revises its data for prior years, which has affected some of the earlier fuel use and population numbers. Second, in response to feedback on the 2013 report, this year’s analysis includes natural gas (a major fuel type subject to the tax). This involved combining two StatsCan datasets (Petroleum Products, and Sales of Natural Gas / CANSIM 129-0003), and then converting all fuel figures from volume to energy (terajoules), to enable comparability (since natural gas has very high volumes per unit of energy compared to other fuels, due to its gaseous state). Statistics Canada confirmed this approach.
Finally, a few values were missing for certain petroleum products for select months from March-June 2013, due to recent changes in data sharing agreements between Statistics Canada and particular companies (suppressed to meet confidentiality requirements). The data for these missing months was estimated, using standard methods, in consultation with Statistics Canada. For the total fuel sales figure for 2012-13, the missing values were estimated using a Kalman filter algorithm. This is a standard method for treating missing values in time series analysis (reference: McLeod, A. I., H. Yu, and E. Mahdi. "Time series analysis with R." Handbook of statistics 30(2011)). For the sales of specific types of fuels, the approach used was to distribute remaining values from the total according to recent relative proportions. Kalman filter and other methods were tested as well, for sensitivity purposes, leading to consistent estimates and end results. Given the limited number of missing values for only a few months of one year, we can have confidence that the final results are accurate with only a very small margin of error.
Back to the 2014 results.
The true story of how B.C’s carbon tax is working
