Appendix B:
Methodology

View the most recent version.

The basic methodology consists essentially of six steps:

(1) taking totals from the CSNEA for jobs, hours worked, and wages and salaries,

(2) further disaggregating these totals using data from the CSNEA for full-time and part-time jobs,

(3) distributing the CSNEA totals for 2000 across occupations, gender and age groups, and immigrant status based on Census data,

(4) building time series from these benchmarks based on movements in corresponding series from the LFS,

(5) smoothing the time series for LFS occupations, while keeping the overall industry group totals,

(6) making limited, final adjustments to data values. This step completes the estimates for total employment of the HRM.

The following describes the methodology and implicit assumptions in more detail.

Step 1: Benchmark totals

The industry group totals on jobs, hours worked, labour income and wages and salaries, including the details by class of worker (employee or self-employed), by full- and part-time status, come from the CPA database which are based on CSNEA principles.

These data did not exactly meet our requirements, however, and a number of adjustments were needed. In particular:

(1) The labour component of mixed income from self-employment is imputed by multiplying the hours worked in self-employment by the average hourly labour income per employee job. This method assumes that the self-employed and paid employees earn the same on average.

(2) The CSNEA jobs data follows the CSNEA version of NAICS, which is simply a special aggregation that defines the working level industries of the Input-Output tables. At this level, detail is available only for two parts of Accommodation Services, Traveller Accommodation (NAICS 7211) and RV Parks, Recreational Campgrounds and Rooming and Boarding Houses (NAICS 7212 + NAICS 7213). The inclusion of Rooming and Boarding Houses requires an adjustment to remove it. This is done using details from SEPH on the industry's share of overall jobs, hours and earnings. These shares are used to adjust both employee and self-employment jobs.

Step 2: Distribution of CSNEA data by full- and part-time status

Data from the CPA database in the CSNEA are used to split jobs, hours worked and income by full-time and part-time status.

Step 3: Distribution of industry totals by occupation

The best information on the occupational distribution of jobs comes from the Census. Step 3 involves using this source to distribute the CSNEA totals by occupation, gender, age group and immigrant status.

The discussion here relates only to employee jobs, as occupational details were not developed for the selfemployed. The Census of Population was used to distribute the industry group totals established in the second step for the year 2000 (i.e., the reference year for the 2001 census).

To develop occupational distributors for the industry group totals, special census tabulations were done identifying persons in the tourism industries who had employment income in the reference year and were not selfemployed. The selected persons were grouped according to their industry and whether they worked mainly full- or part-time during the reference year. For each of these groups the distribution of the (weighted) sample by occupation was determined, as well as the distribution of total hours worked and wage and salary income.

The occupational distributor for hours worked is based on the distribution of total hours (jobs multiplied by average hours worked) across occupations within each industry group.

The occupational distributor for wages and salaries is based on the distribution of total wages and salaries (jobs multiplied by hours worked multiplied by hourly earnings) across occupations in each industry group.

Step 4: Building the occupational time series

Step 4 entails using the corresponding LFS annual average series by occupation, age group and gender to build a time series. Information on immigrant status is not available from the LFS and therefore the percentage distribution from the Census reference year (2000) was maintained for all years. The LFS occupation and industry data matched those used in the Census therefore easing the process. The LFS data were adjusted to the Census level to maintain growth rates between years.

Step 5 and 6: Smoothing the LFS data and manual adjustments

Step 5 entails smoothing (using a four-year moving average) to reduce volatility in the occupational time series found in the LFS. It has been implemented in a way that preserves the overall industry group totals (from the CSNEA) and the occupational distributions (from the 2000 census), as these benchmarks are constraints that must be respected.

A simple moving average is applied to the indicators (discussed in Step 3) used to build the time series on jobs, hours worked and wages and salaries, full- and part-time, by occupation and for each industry. A 4-year moving average was judged to provide the best results overall in terms of reasonableness and consistency, reduced volatility, and minimizing the need for manual adjustments.

The last step in developing the total industry estimates involves manual adjustments when the smoothed series appear out of line, or are notably inconsistent with the hours worked or wages and salaries, or generate erratic movements in the implied average annual hours or average hourly earnings. Results from the feasibility study indicate that the smoothing and manual adjustments have little effect on the general pattern of the occupational distributions. However, they notably reduce the volatility of year-to-year growth rates. Approximately 1% of the data cells were adjusted after the smoothing.