The Canadian Productivity Review

Volume 2007
Number 8

The Comparative Level of GDP per Capita in Canada and the United States: A Decomposition into Labour Productivity and Work Intensity Differences

Main page

More information

PDF version
of the study

About this
series

Previous issues
in this series

The Comparative Level of GDP per Capita in Canada and the United States: A Decomposition into Labour Productivity and Work Intensity Differences

by Jean-Pierre Maynard

Executive summary

This study is the third in a series related to the project launched in 2003 by Statistics Canada's Canadian Productivity Accounts (CPA) to compare productivity levels between Canada and the United States.

In 2005, Statistics Canada's Canadian Productivity Accounts released two studies that, for the first time, examined the comparability of labour productivity levels between Canada and the United States.¹ Previously, Statistics Canada limited comparisons to productivity growth rates. Using analogous sources, concepts and methods to obtain the most comparable measure possible of productivity levels, these new studies found that the Canada–U.S. productivity level difference was lower than normally described.

Neither the Canadian nor the American data used to measure work intensity for this project are the same as those used by many who have conducted Canada/US comparisons of the level of labour productivity. Other studies have used data that were assumed to be comparable—such as data from the Labour Force Survey (LFS) in Canada and those from the equivalent American survey, the Current Population Survey (CPS)—but which are not.

This third study focuses in more depth on the construction of the number of hours worked developed for this project and on the choice of estimates of jobs and population. It describes the reasons that the work intensity measures used in our Canada/U.S. project are superior to alternatives that are readily available but non comparable and therefore inappropriate for studies of Canada/U.S. comparisons of the level of productivity.

Top of Page

This study answers the following questions:

1. What are the reasons for the choice of data to measure the number of hours worked?


2. Why are the estimates of the number of hours worked developed for this study the most appropriate for comparing levels of work intensity and hours worked per job between Canada and the United States?


3. What are the problems with traditional data sources that make them inappropriate for comparisons of levels?


4. What is the degree of error that is made if a study relies on alternate but easily accessible labour force sources to compare levels of productivity and work intensity between Canada and the United States?

Using a framework to analyze differences in real GDP per capita that decomposes the latter into differences in labour productivity and work intensity, this study quantifies the errors committed when alternate, easily accessible but non comparable data sources are used in order to compare the sources of differences in GDP per capita between Canada and the United States.

This table presents the results of the decomposition of the Canada–U.S. difference in GDP per capita for the year 2000 using two inappropriate measures that have been occasionally used for Canada/U.S. comparisons. The first inappropriate measure (line 1) uses estimates of labour input developed by the productivity program of each country to measure the growth in labour productivity. Note that the primary objective of these programs is to estimate productivity growth and not the level of productivity relative to other countries. The second measure (line 2) uses data coming from the monthly household surveys of the two countries—the Labour Force Survey (LFS) in Canada and the Current Population Survey (CPS) in the United States. The third measure (line 3) makes use of the data on labour inputs generated in the Statistics Canada project that developed comparable data to be used to estimate the relative level of Canada–U.S. productivity.

Top of Page

Using the year 2000 as an example and the same GDP per capita measure for the three sources of components, this study shows the crucial importance of using comparable measures, which were developed by harmonizing concepts and coverage and by adjusting data to consider differences in collection methods and in data accuracy. The appropriate comparison (line 3) that uses comparable data shows that labour productivity contributes much less to GDP per capita differences than the two inappropriate techniques.

Measure number 1: Problem with hours per job

The first inappropriate measure uses the levels of the number of hours worked and the number of jobs derived from the official measures used to estimate labour productivity growth in both countries. Using this comparison, 70% of the 20% gap in gross domestic product (GDP) per capita in favour of the United States in 2000 can be attributed to Canada's weaker level of labour productivity. The correct measure (line 3) indicates that only 35% of the gap is due to lower labour productivity.

In general, both countries produce detailed estimates of the number of hours worked by estimating the number of jobs and the annual number of hours worked per job. The number of hours worked is obtained by multiplying these two elements.

The Canadian Productivity Accounts rely mainly on a household survey, the Labour Force Survey (LFS), to estimate employment; in the United States, the starting point for constructing these same estimates is an employer survey, the Current Employment Statistics (CES). Given that this survey has only partial coverage (does not cover, for example, farms and self-employed workers), the Current Population Survey (CPS) estimates are used to complete the coverage. Our assessment is that, based on conceptual, coverage and accuracy criteria, these two measures of employment are appropriate for comparing employment levels between the two countries.

The problem with measure number 1 arises because the estimates of hours per job are derived from different types of surveys that in each country yield quite different estimates of hours worked per job. The Canadian Productivity Accounts rely on hours actually worked collected from a household survey—the LFS; on the other hand, the Bureau of Labor Statistics (BLS) instead uses the hours paid collected from its survey of employers. Although the estimates of hours paid are then converted by the BLS into hours worked by excluding hours of paid leave (vacation, holidays, sick, etc.), these two approaches produce results that are not comparable because household and employer surveys produce estimates that differ in a systematic way.

Top of Page

As part of this project, the Canadian Productivity Accounts conducted a comparison of the estimate of hours worked per job using household and employer surveys in each country. The results from similar surveys were compared across countries (household survey in Canada to household survey in the United States; employer survey in Canada to employer survey in the United States).

The comparison for household surveys made use of a similar methodology to adjust these data for the bias associated with household surveys that do not take into account statutory holidays when extrapolating data from a survey reference week to other weeks in the month. The Canadian Survey of Employment, Payrolls and Hours was used (a survey of employers that gathers data on hours paid or on the regular hours of workers) to estimate the Canadian hours worked following the BLS method that uses an employer survey. This result was then compared to the average hours derived by the BLS from their employer survey.

Each comparison indicated that Americans worked on average one hundred hours more than their Canadian counterparts, annually. But in both countries, the estimate derived from the employer survey was always below the estimate derived from the household survey. The hours worked per job estimated from the data collected from employers produced a level about 8% lower than the level derived from the household surveys after adjustment for statutory holiday bias. This demonstrates the direction and size of the error that occurs when a household survey in Canada is compared to an employer survey in the United States, as is done for measure number 1.

There still remains the issue of whether hours worked for comparisons of levels should be estimated from household or employer surveys. Various studies conducted in several countries, including Canada and the United States, have compared the estimates of hours worked collected from households using a daily survey of time use—in theory the best approach for collecting this information—to the estimates derived both from employer and labour force surveys. The estimates derived from the time-use surveys suggest that labour force surveys provide the most accurate estimates of hours per job. Therefore, these are the estimates that have been adopted in our Canada/U.S. comparison.

This first example demonstrates that the source of the data on hours worked per job is especially important in order to attribute the origin of GDP per capita differences to labour productivity or to hours worked per capita. Comparing hours worked estimated from a survey of employers to those obtained from a household survey has the potential to overestimate the impact of productivity differences on GDP per capita differences between Canada and the United States.

Top of Page

Measure number 2: Sources of labour intensity

The second inappropriate measure (line 2) compares the levels of the number of hours worked, the number of persons employed and the civilian population of working age outside institutions obtained directly from household surveys in both countries. For this comparison, the 20% difference in gross domestic product per capita in favour of the United States in 2000 is divided almost equally between labour productivity (-11%) and work intensity (-9%). As was the case with measure number 1, this one also attributes more importance to differences in labour productivity than the estimate that our Canada/U.S. project yields (line 3).

The differences with our reference measure originate mainly, in this case, with the absolute measures: the number of jobs and the working age population for the United States.

Although, at first glance, Canada's Labour Force Survey (LFS) and the U.S. Current Population Survey appear to provide fully comparable estimates, a more detailed analysis of these two surveys reveals unsuspected and quite substantial differences due to data coverage. These differences are enough to compromise the use of these surveys for direct comparisons of levels of jobs—though not for hours worked per job when comparable methodology is applied to each survey.

While both countries use similar questionnaires, the statistical agencies on opposite sides of the border do not have access to a similar method to calculate the survey frame. In Canada, the demographic weights of the Labour Force Survey are recalibrated every five years using a five-year census, while in the United States, this recalibration occurs only once every ten years. In addition, Canada's recalibration results in an historical revision of the LFS estimates to eliminate any break in the series. In contrast, in the United States, this exercise leads to significant breaks in the Current Population Survey (CPS) series, the most recent having occurred in 2000 and 2003.

Top of Page

Added to this statistical problem is the much higher proportion of illegal immigrants in the United States, whom Bureau of Labor Statistics² analysts suspect are somewhat reluctant to respond to the CPS survey. However, legislation requires employers to report all of their employees annually to unemployment insurance officials and this approach appears to provide a better estimate of illegal immigrants.

In the United States, it is mandatory to have a social security number in order to obtain a job. It is the data from this file that are used as an annual benchmark for the employer survey (Current Employment Survey [CES]), which would explain why exercises to reconcile the two surveys indicate a substantial under-counting of jobs in the CPS compared to the CES. This problem was particularly evident between 1996 and 2003. Corrections made to the population estimate projection model by the U.S. Census Bureau have made it possible to narrow considerably the differences in job estimates between the two surveys since 2003.

As a result, data from the U.S. household survey (CPS) frequently suffer from a problem of underestimating the levels of jobs and the working age population. It also suffers a problem since it only partially revises its series when benchmarking to the Census and this causes breaks in its historical series. These two problems make using job estimates from this survey inappropriate for Canada–U.S. comparisons.

Measure number 3: Reference measures from the Canada/U.S. project for comparing levels

Since the last historical revision of the National Accounts, the Canadian Productivity Accounts (CPA) have developed a measure of the number of hours worked that can be used to measure both the growth and level of labour productivity. This is why Canadian estimates of the number of hours worked and the number of jobs in measure number 3 correspond to the estimates published by the CPA.

Top of Page

In their project to compare Canada–United States productivity levels, analysts with the Canadian Productivity Accounts selected their U.S. data sources to be comparable with the Canadian CPA data.

For several years, the Bureau of Labor Statistics' productivity program has also produced a level of employment that corresponds to the National Accounts concept, which covers the entire American economy and represents the most reliable level of employment that can be developed for that country. These are the estimates derived from their employer survey.

However, there is a problem of comparability with respect to hours per job as described above. As part of the Canada/U.S. comparison project, analysts in the Canadian Productivity Accounts produced estimates of hours worked per job using the Current Population Survey and a similar methodology used for Canadian data to account for holiday bias. It is these estimates that were used to compare the sources of differences in the level of gross domestic product (GDP) per capita.

Lastly, the population estimates used in this article are based on the concept of resident population. This concept is the one used in international GDP per capita comparisons. It is also important to note that it is the U.S. Census Bureau that produces these estimates using a revision procedure that avoids historical breaks.

Although there are differences in the methodologies used by the two countries to produce hours worked estimates that enter into measures of the growth in productivity, as long as these differences remain constant, the accuracy of comparisons of growth rates in the two countries will not be greatly affected. However, these differences in methodology make comparisons of productivity levels more difficult and some care should be used in interpreting and using the data that have been used for comparisons of growth rates. In order to obtain more accurate estimates of productivity levels in Canada relative to the United States, effort is needed to harmonize data sources and methods.

1. Baldwin et al., 2005; Baldwin, Maynard and Wong, 2005.


2. See Nardone et al. Examining the Discrepancy in Employment Growth between the CPS and the CES, FESAC, October 2003.

View the publication The Comparative Level of GDP per Capita in Canada and the United States: A Decomposition into Labour Productivity and Work Intensity Differences in PDF format.

You need to use the free Adobe Reader to view PDF documents. To view (open) these files, simply click on the link. To download (save) them, right-click on the link. Note that if you are using Internet Explorer or AOL, PDF documents sometimes do not open properly. See Troubleshooting PDFs. PDF documents may not be accessible by some devices. For more information, visit the Adobe website or contact us for assistance.