Now that Statistics Canada has made its Cansim database publicly available, Canadians who are interested in public policy can look up the relevant data on their own. But it’s not always clear just what the relevant data are, nor how we should interpret them when they contradict each other.
Employment data are a case in point. The Labour Force Survey (LFS) release makes headlines on the first Friday of every month (the headline numbers are available in Table 282-0087), but it’s not the only source for jobs data.
In the United States, data from the Current Population Survey – its household survey which is the counterpart to our LFS – are published simultaneously with data from its establishment survey of payrolls. Since the number of establishments surveyed is seven times larger than the number of households surveyed by the CPS, the payroll numbers are widely believed to be more reliable than the CPS. And so it is the payroll numbers that make the headlines in the United States.
That’s not the case here – the LFS survey of Canadian households has the media stage to itself. But Statistics Canada also has its own establishment survey, one that is even more reliable than the one in the U.S. Instead of relying on a large random sample of employers, the Survey of Employment, Payrolls and Hours (SEPH – headline numbers are available in Table 281-0025) is based on administrative data from the entire population of employers.
Unlike the LFS, there is no sampling with the SEPH, so there is no sampling error. The price of such precision is timeliness: the SEPH numbers come out two months after the LFS. So when the SEPH numbers are finally released, they are invariably given the level of scrutiny that is reserved for two-month-old news. While U.S. analysts give priority to their payroll numbers, our establishment data are treated as an afterthought.
This doesn’t mean that we should ignore the LFS numbers: it is the only survey that links employment to demographic data. And there are some important conceptual differences between the LFS and the SEPH data: for example, the LFS includes the self-employed, and the SEPH does not.
The recent differences between the two sets of employment numbers are enough to change the tone of the labour market narrative of the past few months. The story we've been telling from the LFS data is that employment growth has been flat since mid-2011. But a story that takes into account the SEPH would revise that LFS-based narrative towards one in which employment growth has continued, but at a slower rate.
Now that the Cansim data are freely available, we should make the most of them. In my view, the SEPH employment data have been undervalued in the public policy arena, and they deserve more attention.
(A more detailed version of this post is available at Worthwhile Canadian Initiative.)
For more of Stephen Gordon's recent posts, click here.
Follow Economy Lab on twitter