Statistics Canada
www.statcan.gc.ca
Common menu bar links
Main article
- Introduction
- Revision process
- Impact of revisions on labour productivity
- Conclusion
- Statistical Tables
1 Introduction
In this paper, we compare recent revisions to the labour productivity estimates in Canada and the United States 1 . These regular revisions to preliminary estimates extend back four years.
Revisions to Canadian labour productivity estimates have been made in June 2008 and in December 2008 in order to incorporate the latest available gross domestic product (GDP) estimates published by the National Economic and Financial Accounts and the annual benchmarks on hours worked. These revisions relate to the last four years (2004 to 2007). Recent productivity data are produced on the basis of preliminary GDP and hours worked estimates, which are eventually revised when additional and more precise information sources become available for the National Accounts.
In August 2008, revisions were also made to labour productivity in the United States for the three last years, from 2005 to 2007.
2 Revision process
The estimates of labour productivity (output per hour worked) that are produced by the Canadian Productivity Accounts are subject to two types of revisions. Please take note that revisions for this year are only of the first type.
Revisions of the first type are a series of annual revisions in the gross domestic product (GDP) that go back over a four-year period (Statistics Canada, 2008). With this revision cycle, a preliminary estimate of GDP first released in year t is revised annually over the four subsequent years (t+1 to t+4) as more detailed and accurate data become available to the System of National Accounts (SNA).
When first released, estimates of GDP at the industrial sector level come from projecting past estimates using a small number of readily measured series (for example, the GDP in Taxi and Limousine Services is projected off the Survey of Employment, Payrolls and Hours [SEPH] estimate of employment growth in these industries). The industry estimates are gradually supplemented by far more detailed and accurate data that are obtained from surveys such as the Annual Survey of Manufactures and the Unified Enterprise Survey, and from administrative tax records that become available after a lag of one or two years. Preliminary estimates of GDP that are calculated from final demand are also projected the first time from sources that are eventually replaced by more comprehensive information.
In addition, the labour productivity estimates for year t are revised in year t+1 as new information becomes available to improve the first estimates of employment and hours worked that are made using the Labour Force Survey (LFS) and SEPH. These revisions improve the estimates of hours worked because more precise measures of holidays and other non-random events are used (Maynard 2005). Revisions also occur if the employment estimates for the non-commercial sector obtained from the Public Institutions Division and produced with SEPH are revised, since the business sector estimate is obtained residually after removing the non-commercial sector from the total economy.
Revisions of the second type occur less frequently, about once or twice every 10 years. Historical revisions of the SNA are occasionally carried out to eliminate breaks in some series, to modify classification standards (for example, the movement from the Standard Industrial Classification to the North American Industry Classification System) or to introduce conceptual and methodological changes. These revisions of the second type occur when the SNA updates the method used for measuring certain industries, sometimes because of changes in the international standards to which it adheres (SNA93). For example, in 2001, Statistics Canada included software expenditures as investment for the first time 2 . In addition, the adoption of Fisher chain indices instead of fixed base Laspeyres indices has introduced revisions to GDP and, therefore, a historical revision to the labour productivity estimates.
Historical revisions also occur in the employment and hours worked data when the LFS is occasionally re-benchmarked against data from the most recent Census of Population.
3 Impact of revisions on labour productivity
In this section, we compare recent revisions to the labour productivity estimates for the business sector in Canada and the United States.
Revisions to Canadian labour productivity estimates have been made in June 2008 in order to incorporate the latest available gross domestic product (GDP) estimates published by the National Economic and Financial Accounts 3 and in December 2008 when revised annual benchmarks on hours worked were published. These revisions relate to the last four years (2004 to 2007). The revisions to estimates of labour productivity, GDP and volume of hours worked, which extend back to 2000, are presented in Tables 1 to 3 . These tables show a picture of the evolution of the estimates over the last seven revision rounds since 2002. It should be noted that the revision cycles of the most recent estimates are not yet completed. This is the case for the 2005-to-2007 estimates for Canada and the 2006 and 2007 ones in the United States. The estimates produced during the first four-year revision cycle appear with the footnote marker “1.” Other revisions reflect revisions of the second type outlined above.
Table 4 of this paper shows the impact of revisions on the respective labour productivity performance of Canada and the United States for two periods—1981 to 2000 and 2000 to 2007 4 . It should be noted that the latter period is less than a business cycle in length and covers the years since the end of the previous peak in productivity growth that was observed in 2000. It also corresponds to a period (except for 2000 to 2003) when only preliminary estimates of GDP are available. The former period, however, contains mainly estimates that are past the preliminary revision cycle. It essentially covers two business cycles and therefore provides a better comparison of differences in long-term trends between Canada and the United States 5 . Productivity estimates of short-term changes are generally more volatile than estimates of changes over the long term.
3.1 Impact of Canadian revisions
In general, the revisions of the Canadian productivity figures for the period from 2004 to 2007 that were published in The Daily on December 10, 2008, resulted in almost no change from previous estimates. For businesses on the whole, the revisions had the impact of reducing the annual rate of growth in Canada’s labour productivity for 2005, and increasing it for 2004, 2006 and 2007. The magnitude of the revisions ranged from a 0.3% decline in 2005 to a 0.3% gain in 2006.
For this entire period, average annual productivity growth was revised upward by 0.1%. It should be noted that calculations of the productivity growth rate and its related variables are now based on index numbers rounded to three decimal places. On CANSIM, these calculations would still be based on index numbers rounded to one decimal place.
3.2 Impact of American revisions
On the other hand, the annual growth rate of labour productivity in American businesses that was published in News of the Bureau of Labor Statistics on August 8, 2008, was revised downward for each of the last three years, from 2005 to 2007. For this whole period, the revised estimates show that productivity increased only 1.2% on average, 0.3 percentage point less than the previously published estimates indicated.
Thus, the average annual rate of productivity in the United States reached 1.8% in 2005 (instead of 2.0%), 0.9% in 2006 (instead of 1.0%) and 1.5% in 2007 (instead of 1.9%). Annual productivity growth in the United States has slowed gradually since 2002 (the year where it peaked at 4.1%), before rebounding slightly in 2007.
Since 2002, the United States has systematically revised downwards
its preliminary estimates of labour productivity. On average, revisions have
reduced productivity growth by
3.3 Analysis of the gap between Canada and the United States
With those revisions, average annual productivity growth in American businesses over the period from 2004 to 2007 was 1.4% (compared to 1.6% before revisions), the same growth rate as in Canadian businesses. Both the old and the new estimates of growth in labour productivity and hours worked are presented in table 4.
However, there was a substantial difference in average annual productivity growth between Canada and the United States from 2000 to 2004, with U.S. productivity growing almost five times faster than Canada’s.
Between 2000 and 2004, the average annual growth in U.S. productivity was 3.3%, compared to the 0.7% posted in Canada.
Over the 2000-to-2007 period, the average annual growth rate was 1.0% in Canada and 2.5% in the United States; U.S. productivity growth was two and a half times higher than that of its Canadian counterpart. During that period, real GDP growth was similar on both sides of the border, but hours worked increased at a faster pace in Canada. More precisely, GDP growth in Canada was 2.5% on average per year from 2000 to 2007, while the hours worked increased by 1.5%. In comparison, the U.S. GDP grew by 2.5% on average, while hours worked remainded unchanged during the same period.
It should be noted that the period from 2000 to 2007 covers much less than a full business cycle. In the early 1990s, Canada also lagged behind the United States in productivity growth, particularly during the recessionary period, but ended up at approximately the same point by the end of the decade.
It is also important to note that the annual productivity differences reported over the period from 2000 to 2007 are based on preliminary data that are still subject to revision. Since 2002, the United States has also systematically revised downward its preliminary estimates of productivity. It is important to take into account this additional margin of error when analysing data of recent years.
Since 1998, the Canada–United States gap has generally shrunk, following revisions to the preliminary data. The main revision made to the Canadian productivity estimate in 1999 is almost entirely due to revisions in GDP. Almost half the revision in 2000 comes from this source 6 . During this period, two changes were made to the Canadian System of National Accounts, which increased the rate of growth of output and therefore of labour productivity. First, software expenditures were capitalized (the United States had introduced this in 1999). Second, new surveys were gradually introduced, which were associated with the Project to Improve Provincial Economic Statistics whose economic survey coverage had been extended 7 . During this period, the productivity program also revised its estimate of hours worked, downward 8 .
Over a longer period (1981 to 2000), there is a small gap in productivity growth between Canada and the United States (0.3 percentage point per year), some of which may arise from slightly different methods used to calculate the growth in labour inputs (Maynard 2007). For this period, productivity has grown at an average annual rate of 1.6% in Canada compared with 1.9% in the United States.
4 Conclusion
All things considered, the 2004-to-2007 revisions of gross domestic product (GDP) in Canada (and the 2005-to-2007 revisions of GDP in the United States) resulted in a slight narrowing of the gap in productivity growth. Since the most recent revisions applied only to the period after 2000, they have had little effect on Canada–United States differences over the last two decades—from 1981 to 2000. During this period, Statistics Canada’s estimates of productivity growth have consistently shown an annual gap of 0.3 percentage point between Canada and the United States. For a more extensive discussion of the significance of the difference and the causes behind it, see Statistics Canada (2007).
5 Statistical Tables
