International productivity comparisons

View the most recent version.

Data that may be fit for one purpose—that meet acceptable quality standards in one area—may not be for others. And statistics may be developed for one purpose but users may begin to employ them for other purposes—for which they were not designed and for which they may be less than ideal.

The evolution of the Canadian productivity program provides an example of just such a transition. Statistics Canada's productivity accounts were originally developed to provide information on productivity growth rates in Canada—first with regards to labour productivity and then to multifactor (what academics often refer to as total factor) productivity. In a world of increasing globalization, demands of users for international comparisons have increased. Providing estimates for international comparisons that meet acceptable quality standards poses particular challenges.

The productivity program at Statistics Canada first focused on providing information products that compare Canada–United States productivity growth rates, choosing U.S. estimates that are closest to the Canadian ones. Despite differences between the two countries in sources used, these differences are sufficiently stable over time that they generally do not provide a major problem for comparisons of Canada–United States growth rates.

But the summary statistics produced by the official productivity programs of the two countries turn out to be less than ideal for analysing differences in productivity levels. Analysts have used data that are employed in the growth programs to generate cross-country comparisons of levels.

While the data that were being used for this purpose were not generated for purposes of estimating differences in levels, statistical agencies have to respond to users needs since relevance is an important aspect of quality of product.

In Canada, users have requested guidance on the quality of Canada–United States productivity comparisons. Recently, Statistics Canada commenced a set of studies that examined alternatives that can be used to estimate the level of relative productivity—both labour and multifactor productivity. Statistics Canada found that despite the relative similarity in the statistical systems of the two countries, improved harmonization of data sources and methodology was required in order to produce better estimates of the relative level of Canada–United States productivity.

Our studies here pointed out several problems with many previous attempts to compare Canada– United States levels of labour productivity.¹⁶ First, studies were sometimes not using measures of gross domestic product (GDP) that were comparable. GDP is measured at market prices, at basic prices and at factor cost. And the level of GDP that is produced by these estimates can vary by up to 16%. Secondly, comparisons of levels of GDP across countries need to take into account differences in price levels if relative values of output are to be transformed into relative levels of real output. For this purpose, purchasing power parities (PPPs) are necessary and the existing PPPs are sufficiently imprecise as to produce estimates of relative levels of output with quite large confidence intervals around them. Finally, and most importantly, obtaining accurate estimates of relative labour input provide particular challenges. Differences exist in the way that labour input is calculated in the official productivity programs of both countries that have led to a substantial downward bias in the relative Canadian level of labour productivity when it is derived from the 'official' sources of labour productivity from each country. The estimate of total hours worked comes from the product of number of jobs multiplied by hours worked per job. The estimate of hours worked per job that is derived from a labour force (household) survey is generally higher than that derived from an employer survey. The Canadian productivity program relies on the former while the U.S. productivity program relies on the latter. When comparable sources are used for both countries (whether they be household or employer surveys) in both countries, the relative labour intensity in Canada increases by between 5% and 10% relative to the estimate derived from each country's official estimates used in the productivity growth programs.

 

16. See Baldwin et al. 2005 and Maynard 2007b.