July 2014 LFS: An alternative view. Illustrating two critiques of the so-called ‘jobs report’

Chart 1 Type of work by class of worker (usual hours)
Chart 1 Type of work by class of worker (usual hrs) Jan2003-Jul2014

Source: CANSIM Table 282-0019 Labour force survey estimates (LFS), by usual hours worked, class of worker, North American Industry Classification System (NAICS) and sex, unadjusted for seasonality, monthly (persons unless otherwise noted), Statistics Canada

By now it should come as no surprise that a guesstimated June 2014 23,400 self-employment gain was followed by  a July 2014 29.200 self-employment loss – with the usual caveat that all these monthly figures are meaningless as they’ll be revised in the coming months.

Following the July 2014 Labour Force Survey (LFS) release this morning, which reported no new jobs yet a drop in the unemployment rate (as more Canadians supposedly gave up looking for work), media focus turned to the recent full-time / part-time employment trend.

Since self-employment was touched on last month, and part-time employment the month before last, why not combine the two? This would give a better picture of employment the Canadian economy created, versus employment struggling Canadians have increasingly tried – but (so far) remarkably failed – to create for themselves in the lengthening shadow of the Great Recession.

US labor market data

The US Bureau of Labor Statistics (BLS) monthly jobs report derives its two key stats from two different surveys. The unemployment rate is derived from the labor force component of its household survey, the Current Population Survey (CPS). The jobs figure is derived from its employer payroll survey, the Current Employment Statistics (CES).

To aid with comparability between the two labor market data sources, BLS publishes monthly Table A-8. Employed persons by class of worker and part-time status. It separates the CPS labor data into agricultural and non-agricultural employment, as the CES excludes the agricultural sector.

It also separates self-employment in the CPS labor data into incorporated  and unincorporated, lumping the former in with ‘Wage and salary workers’ in its reported figures. On the one hand, this  may make sense given that an incorporated business with employees could be in-scope for CES; on the other, such a business is unlikely to meet other criteria for CES inclusion (example, minimum 1 full year in operation as an employer).

While the CPS does not collect monthly earnings info for the self-employed, the US Census Bureau (which conducts surveys on behalf of BLS) does collect the information on an annual basis.

Canadian labour market data

While Canada does have a separate payroll survey, the Survey of Employment Payroll and Hours (SEPH), both the unemployment rate and jobs data commonly cited in the media are derived from the household LFS. That’s because unlike the BLS, StatsCan does not release the results of its household and employer payroll surveys on the same day; SEPH is released two months later, which is two months too late.

Compounding the problem is that StatsCan does not publish CANSIM tables like BLS Table A-8 to facilitate comparison between the two labour market data sources. It’s possible to create such a table with the LFS public use microdata file (PUMF), but it’s a PitA – assuming one has access to the historical series (which we do; perhaps a future post/update).

Neither the LFS nor any currently active StatsCan survey we’re aware of collects self-employment earnings info.

Also, for reasons unknown and explained, StatsCan defines the full-time cut-off as 30 hours per week, whereas the BLS defines it as most of the industrialised world does – 35 hours.

Excluding self-employed, adjusting for hours

What would be interesting to see is Canadian employment type (full, part-time) by class of worker (employee, self-employed), adjusted for the normal (35 hours) cut-off. It’s unclear how accurately the BLS approach of counting incorporated self-employed with payroll-employed is. And there’s no readily available CANSIM table that includes both detailed hours and detailed class of worker. So the best one can do is separate out all self-employed to make the household survey data more comparable to that of the payroll survey, as well as adjust for a usual or actual 35-hour full-time cut-off.


As illustrated in Chart 1, all the labour market trends prior to October 2008 (when the Canadian LFS started capturing recessionary job losses) were inverted thereafter. As the recession hit, part-time self-employment rose, regular part-time employment and full-time self-employment stagnated, while regular full-time employment dramatically declined.

Those trends have largely remained inverted since. Regular full-time employment recovered starting mid-2009 through late 2012, but has flat-lined thereafter. Regular part-time employment growth has been modest over the same period, and full-time self-employment has significantly declined. So what’s been the only Canadian labour market bright spot? Answer: Part-time self-employment. Since StatsCan doesn’t collect earnings info from the self-employed, it’s a safe bet that LFS-reported wage growth does not accurately reflect this labour market ‘bright spot’.

Either StatsCan needs to start collecting self-employment earnings data, or it needs to figure out how to move the SEPH release up to the same date as the LFS. Continuing to call the LFS the ‘jobs report’ not only confuses comparison with the US actual payroll jobs report, it just doesn’t make sense on the face of it.

Note: Chart 1 data was seasonally adjusted using TRAMO-SEATS from Banco de España.

Update 12/07/2014
Apparently StatsCan pulled the July 2014 LFS report due to some unspecified error. That should be the least of Canadians’ concerns with the survey’s results…

Leave a Reply

Your email address will not be published. Required fields are marked *