|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||
Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.
|
The Canadian Productivity Review Volume 2006, Number 4 Producing Hours Worked for the SNA
in Order to Measure Productivity: the Canadian Experience
|
|
What is produced in CanadaSince 2001, the Canadian Productivity Accounts (CPA) have produced data on the number of jobs and hours worked by category of worker for the 286 industries of the Industry Accounts, for the 10 provinces and 3 territories of the Canadian North. Before the incorporation of the SNA (1993), the program aimed essentially to measure growth in labour productivity. In the 1990s, it became clear that our clients also used these data to compare productivity levels. This realization led to the development of the current database intended to estimate both trend and level. Methodology MethodologyCanadian methodology was developed in the 1960s by analysts in Statistics Canada's Productivity Accounts group. Over the years, improvements have been made to both the methodology and the source data, but the basic principles have remained essentially unchanged. In short, the CPA produces detailed estimates of the hours worked by estimating the number of jobs on the one hand1 and the average annual hours per job on the other. The volume of hours worked is obtained by multiplying these two components together. (See the equation below). When constructing estimates of the labour input, the CPA has three objectives:
The use of a two-stage strategy gives us more latitude in the choice of the source survey data necessary to reach these objectives.
J = Number of jobs
Vh = Volume of hours worked. Where c = class of worker, i = industry and r = region. In practice, there is no single source in Canada for estimating a labour input that corresponds entirely, both conceptually and with respect to coverage, with the international standards set out in the SNA Manual of 1993. Canadian data on hours worked are therefore derived by integrating the results from several sources: firm-based employer surveys and household surveys, to which are added the results of quinquennial censuses and administrative data. Main data sources for labour inputStatistics Canada collects labour data mainly from the Labour Force Survey (LFS) and the Survey of Employment, Payrolls and Hours (SEPH). The LFS is a monthly household survey providing data on persons, hours worked and payroll. In a specific week every month it collects information on 53,000 households (about 100,000 persons aged 15 and over). The sample is benchmarked to population censuses and inter-censal estimations. It constitutes the Canadian benchmark for the development of the overall measure of labour market. Given its survey frame, this survey is considered the most reliable for estimates at the aggregate level. Its methodology has not changed since 1976 and the comprehensive nature of its questionnaire facilitates a conceptual harmonization of labour estimates with the SNA (1993). This requires that persons with more than one job be captured and, that persons absent from work who were not paid during the week of the survey be excluded. The concept of hours worked parallels the SNA's. Since the LFS is a household survey, the data are not the most reliable source for detailed industry coding.2 In this respect, the SEPH data obtained from an establishment survey are more suitable. A census of the industries covered by this survey is made each month and the industrial coding of the base units — i.e., the establishments — is taken from a business register of firms that uses a standardized classification system (NAICS). This is the common register for all establishment surveys on specific industries. The SEPH collects monthly data on employee jobs and payroll from all establishments in Canada, except agriculture, fishing and trapping, services to agriculture, services to private households, religious organizations and defence services. Since 1998, when remitting payroll deductions, all employers must report the number of their employees and gross payroll, ideally for the last pay period of the month, to the national tax authority (Canada Revenue Agency).3 In light of the available data, our strategy is to capitalize on the
strengths of each survey, while reducing the impact of its weaknesses.
The choice of data sources is determined by their degree of correspondence
with the proposed concept, their degree of accuracy at various aggregate
levels, their extent of coverage, their conformity with sources used to
measure production and their methodological consistency over time (no
historical break). Estimation of hours workedThe estimation of hours worked takes place in four stages:
Estimation of average hoursIn estimating average hours, the CPA relies mainly on LFS data. This survey consists of a series of questions on the weekly work schedule of employed persons. Respondents are first asked about their regular schedule, paid or unpaid overtime and hours lost. Then they are asked to specify the number of hours worked during the reference week. In the event that respondents were absent from work during the reference week, they are also asked to state the main reason for their absence. Hours worked in LFS reference weeks are first adjusted by the CPA to eliminate sporadic events (civic holidays, strikes, etc.). They are then interpolated to produce estimates for all weeks in the year. When they are annualized, the hours lost due to sporadic events identified by the survey as well as those outside of the reference week are systematically reintroduced into the calendar. Lastly, a final series of adjustments is made to account for the day on which each year begins or ends. (See Figure A1). As can be seen in Table A2, if LFS hours worked are not adjusted for sporadic civic holidays, the estimates of the level and trend of average hours worked will be biased.4 Interpolated hours are calculated by region, industry and class of worker. An overall benchmark for average annual hours worked is produced for each region. At the most detailed industrial level, average hours worked are annualized by combining interpolated hours and employment derived from various information sources.5 (See Calculation of industry details). Calculation of benchmarksAt the aggregate level, the LFS is the main source of data for jobs and hours worked per job for each province and territory, as well as for the three labour categories (paid workers, self-employed with paid help and self-employed own account). The results of this survey are used as the core benchmark. To complete the spatial coverage and harmonize the regional labour concept of this survey with the SNA's,6 other sources have been added: the census estimate of inter-provincial flows of persons employed, SEPH administrative data for Canada's northern territories and Aboriginal reserves and data from the Public Institution Division for employees of different levels of government. (See Table A1 for an example). The aggregate payroll data are taken from administrative data for the entire economy. Calculation of industry detailsThe industry breakdown of the initial job and payroll matrices by province and territory is based on surveys of various job categories. The main source for employees is the SEPH, with the exception of industries excluded from this survey as well as construction, retail trade and accommodation and food services,7 whose source is the LFS. (See Step 1 of Table A6). The main source for self-employed workers is Population LFS data that is reconciled with the detailed industrial distribution interpolated from quinquennial censuses. These results are adjusted to special LFS8 sub-totals used as annual benchmarks for each province and category of self-employed worker (employer or own-account) (Step 5 of Table A6). Since the Labour Force Survey uses the Population Census to benchmark its population weights, labour market data collected from the survey are consistent with the Census. Reconciliation of jobs with the Industry AccountsIn the next stage, the CPA reconcile and modify if necessary, the initial job and payroll matrices by comparing them with those in the Industry Accounts. It forces the resulting estimates to provincial benchmarks. It should be noted that the Industry Accounts rely on various data sources to estimate the Input-Output accounts. In particular, they use annual administrative tax data as a benchmark for salaries and wages, whereas industry coding is mainly derived from specific surveys of each industry.9 In the process of reconciliation by industry, the CPA consider known occurrences, such as: opening and closing of establishments, massive layoffs, strikes, power outages, etc. which lead to unusual job fluctuations. Also taken into consideration are retroactive payments, annual bonuses and pay in lieu of notice, which cause sudden fluctuations in compensation (Step 2 of Table A6). The reconciliation involves a multistop algorithm. In the first stage of the reconciliation, the survey payroll is compared to the wages and salaries of the Industry Accounts. If there are large discrepancies, the breakdown of jobs between industries is adjusted. For a given industry, the degree of adjustment for jobs will depend on the relative variance of job and average compensation estimates from the labour data source used to produce the initial estimate of this industry. The more reliable the quality of a cell, the less it will be affected by the reconciliation algorithm. Our variance estimates take into account not only sampling errors, but also non-sampling errors. Once the calibration of job estimates has been completed, the volume of hours is obtained by multiplying the job by the average hours previously calculated (Step 7 of Table A6). Finally, the number of jobs and the volume of hours are forced to their benchmark value. A similar exercise is applied to self-employed jobs via the mixed incomes of unincorporated businesses contained in the Industry Accounts (Step 5 of Table A6). In the Canadian Economic Accounts, the construction industry includes both contracted-out and own-account activity.10 The employment level of own-account construction is estimated from the available or projected payrolls in Industry Accounts. Jobs are then derived from relative salary rates captured by the LFS for professions likely to fall in this category. Estimated jobs are removed from the industry of their employer and added to construction. In 2002, own-account construction represented about 13% of the hours worked in total construction activities in Canada (Step 3 of Table A6). Data for jobs, hours worked and compensation are available for SNA sectors, as well as for industries. For this purpose, the non-commercial sector is based on job statistics in the Public Institution Division and on the imputation of work in non-profit organizations serving private households. The business sector is obtained residually (Step 4 of Table A6). Strengths and obstaclesA comparison of our results with a 1998 time use survey in Canada indicates that our method of annualizing hours worked is reliable. The derivation of hours lost also corroborates the validity of our estimates, since hours lost due to annual vacations and civic holidays are consistent with Canadian labour legislation (See Table A3 and Table A4). Since 2001, labour statistics in line with the SNA have been produced by province and territory from 1997. Each region expects to obtain data of a similar quality. Unfortunately, Canada is made up of regions of diverse size and population and the availability and quality of labour force data is relatively poorer for the North and for the Atlantic provinces.
|
|
|
= Average annual hours worked
