Source: CANSIM Table 282-0011 Labour force survey estimates (LFS), employment by class of worker, North American Industry Classification System (NAICS) and sex, unadjusted for seasonality, Statistics Canada
After skimming through half a dozen national media and labour reports following the June 2014 Labour Force Survey (LFS) release this morning, it was surprising / disappointing that not one mentioned self-employment. All led with some variation of “Canada lost 9,400 jobs.” As indicated in Table 2 of the release, Canada supposedly lost an LFS-estimated 32,800 payroll jobs, the more modest 9,400 figure tempered by the estimated 23.400 gain in self-employment — with the usual caveat that all these monthly figures are meaningless as they’ll be revised in the coming months.
There are more than a few reasons to be wary of a ‘jobs’ report that’s increasingly being affected by self-employment noise.
Effect on US, Canadian employment comparability
The problem is exacerbated when Statistics Canada’s LFS job figures are compared with the US Bureau of Labour Statistics (BLS) jobs data. While the BLS does have a labour force survey – the Labor Force Statistics component of the Current Population Survey – it only uses it to report the unemployment rate. The employment figure BLS releases on the same day as the unemployment rate is obtained from a separate employer payroll survey – Current Employment Statistics – which excludes both farm and self- employment.
For reasons not very well justified, StatsCan’s payroll survey – Survey of Employment Payroll and Hours – lags two months behind the LFS. Since media and labour groups want to report both the unemployment rate and employment figure at the same time, and the LFS comes out first, we ended up with the unfortunate situation we have now. To repeat StatsCan’s LFS employment estimate is not comparable to that reported by the US BLS – since the LFS estimate includes farm and self- employment, it will always be (relatively) higher.
Self-employment should be excluded
Another reason self-employment should be excluded: While StatsCan collects hours worked from those self-identifying as self-employed, it does not collect earnings information from them. The reason’s kind of obtuse: Since the self-employed don’t earn a salary, and could re-invest their earnings into their business instead of retain it as personal income, any such information collected would be inconsistent and unreliable. There’s a fairly obvious solution: For those that identify as self-employed, have the interviewer ask them to provide only the earnings they retained as personal income.
As it stands, there’s no way to determine whether someone identifying as self-employed is earning a decent living, or simply opting to try self-employment for lack of payroll employment opportunities.
Without the benefit of earnings information, one possible – albeit crude – way of trying to assess self-employment quality is by looking at the trend in such enterprises with ‘paid help’, incorporated or otherwise.
Why not simply incorporated alone? Interesting fact: In the age of increasing precarious employment, more workers are employed as ‘perma-temps’. They may be continuously employed, but through temp agencies that provide them only short-term contracts (either with the same or different employers). Pretty much all such agencies offer temps two payroll options, the more attractive – being paid as an independent contractor – only available to those who (self-) incorporate. All of which is to say, it’s unclear how many incorporated self-employed without ‘paid help’ are actually working at their own enterprise, or working as contracted perma-temps.
As illustrated in Chart 1, the volatile overall self-employment trend that’s seen little change in the average self-employment share of the labour market hides an underlying trend. Based on our crude measure, self-employment quality has deteriorated as the share of self-employed individuals rose significantly as the self-employed who could afford / take on paid declined since the Great Recession.