There's something wrong with Canada's job market statistics
Every month, Statistics Canada releases their Labour Force Survey, a barometer of the health of labour markets at the provincial and metropolitan levels. Last week, the April numbers were released, showing unprecedented job losses in some of Canada's largest cities. What is particularly troubling is that it is not clear what is causing it.
The monthly census metropolitan area (CMA) data in the Labour Force Survey (LFS) stretches back 170 months to 2001. Due to sample size, the monthly CMA data is on a three-month moving average, so the April data for each CMA contains data from February, March and April. The data, collected on a survey basis, includes estimates on the number of employed persons and the size of the labour force (which is roughly equal to the number of employed persons plus persons without jobs but actively looking for work). The LFS data is available in both a seasonally-adjusted and a raw form, with the seasonally adjusted figures typically cited, including in this piece.
April’s numbers are rather dramatic, with five CMAs (of 33 total across Canada) in three provinces seeing unusually large declines in both employment and the size of their labour forces:
Halifax lost 2,500 jobs in April, its worst showing on record (the old record was 2,200 jobs lost, set the month before). Their labour force declined by 1,900, the third worst (of 170) showing on record (tied with March 2015).
- Oshawa shed 3,000 jobs (second worst month on record) and saw their labour force shrink by 2,400 (fifth worst on record).
- Kitchener-Cambridge-Waterloo had a record setting month for both job loss (6,200) and persons leaving the labour force (5,000).
- London also saw a record 5,000 people leaving the labour force, but their employment decline of 3,700 was only the fifth worst on record.
- Vancouver had their third worst month ever for employment growth (-8,900) and their second worst month ever for labour force growth (-8,400). British Columbia as a whole had its worst ever month for employment (28,700 jobs lost). The provincial level data covers 471 months, going back to 1976.
I used absolute declines, though the results look similar if percentage declines are used instead. We have five different CMAs each with employment and labour force results among their worst five ever. This list excludes Barrie, which lost 1,800 jobs (fifth worst showing ever) but had a modest 500-person decline in the size of their labour force.
There does not seem to be any obvious economic reasons for large job and labour-force losses that would be isolated to these five cities. Any combination of economic explanations, from Target leaving Canada to Future Shop closing to the re-tooling of Chrysler’s Windsor plant are in the wrong areas geographically and are too small to account for such substantial job market declines.
I could think of three possible non-economic explanations for this phenomenon, though none of them appear to hold either:
Perhaps such large drops are not that unusual
My first thought that was that this phenomenon of multiple cities simultaneously experiencing labour market declines has occurred before but I simply had not noticed. That does not appear to be the case. In the 170 months with records, only one month saw even three CMAs experience similar declines. In December 2010, Barrie, Sudbury and Abbotsford all experienced significant declines in both the size of their labour force and the number of employed persons. Interestingly, none of the months in or immediately following the recession saw more than two CMAs simultaneously have “bad” months. One important caveat is that the job numbers for past months may have been revised, while the current ones have not undergone a revision process.
Perhaps the January numbers were off
The April CMA figures contain data from February-to-April while the March figures contain data from January-to-March. Any large drops between March to April could be explained by January being dropped out of the sample. That does not appear to explain the declines, with January growth figures for all five cities right around historical averages.
Perhaps the seasonal adjustment was off
Since all the estimates cited were seasonally adjusted, it would be possible to get unusual results if a mistake were made in the adjustment process. This also does not appear to be the case, with the same phenomenon appearing in the seasonally unadjusted data. Furthermore, the seasonal adjustment factor appears similar to that used in past Aprils.
I am stumped. There is no reason why the April employment data should show such a massive decline in the labour market conditions of five cities in three provinces. The only remaining explanation is that these are simply bad estimates, likely due to small sample sizes. It would be unfortunate if this were the case, as this data is reported on with the assumption it is reliable and local policy makers and economic development agencies use the data when making policy. If we are to have smart, efficient and cost-effective public policy, we need to be able to trust that the data is accurate. Given the sheer number of anomalies and last year’s data recall, I cannot put much stock in the Labour Force Survey.
Date: May 13, 2015