The superlatives were flowing last week after Statistics Canada shocked markets and economists with the news that the employers had gone a hiring spree in May.
The gain was officially listed as 95,000, only 100 shy of the biggest increase in 35 years and that included a 6,200 drop in self-employment, making the number of new employees collecting paycheques even higher.
Not only was the gain “stunning,” “spectacular,” a “blow-out” — pick your adjective as many did — it was almost certainly wrong.
As economists noted at the time and continue to do so, there was nothing in the indicators leading into the month to suggest the Canadian economy had suddenly gone from second gear to warp-speed.
Business surveys, including from the Bank of Canada, have detected only luke-warm enthusiasm for adding workers, and wage gains remain modest at just over two per cent over the past year. As well, despite there being significantly more workers, almost all full-time, the number of hours worked during the month actually fell 0.2 per cent.
Statistics Canada has reported wild fluctuations before in its labour market survey, including most recently a 51,000 job gain in February that was followed by a 55,000 loss in March, but the May result was on another scale altogether. There was even a 54,000 gain in youth employment, a category that had seen little movement since the end of the recession.
The most obvious explanation was that the survey of 55,000 Canadian households conducted monthly by Statistics Canada went “rogue” in May.
“That’s what we’re wondering,” says Jimmy Jean, an economic strategist with Desjardins Capital Markets.
“There’s been a number of (staff) cuts at StatsCan. They went out of their way to say this did not affect the employment estimates, but you kind of wonder.”
Asked to respond, Statistics Canada said in an email reply that “there were no budgetary reductions made to the Labour Force Survey program.”
Senior economist Michael Gregory of the Bank of Montreal also questioned whether government cutbacks was impacting how the agency went about collecting data, but added as a former participant in the survey, he has been skeptical for some time.
“It is very much how people respond,” he says. “How many hours did you work this week? There’s no independent verification. The results are not even revised,” as they are monthly in the United States. In Canada, the agency only publishes an annual revision.
To be fair, the agency is up front with its how exacting the results should be interpreted. In April the agency began inserting standard error factors in their tables, and the results Friday mean the actual job gains would have been anywhere between 37,600 and 152,400 with a confidence rating of 95 per cent.
Labour economist Erin Weir says he is not in the practice of questioning the Statistics Canada jobs numbers, but points out that it is a survey subject to the standard problems of opinion polls.
“Even at 95 per cent confidence, the survey will still be wrong one time out of twenty. So, we can reasonably expect a wacky monthly report about once every couple of years.”
Because the variation is so wide, economists say they always take the monthly data with a grain of salt, especially if it’s a surprise. Unfortunately, markets don’t, often reacting to the headline number within seconds — as happened Friday with the loonie rising almost a cent after the 8:30 a.m. release before settling down to close six-tenth of a cent US higher.
A truer picture emerges by looking a three-month or six-month rolling average, which smooths out the monthly peaks and valleys, say analysts.
But even by that measure, Jean says the labour survey has been out of sorts with another survey conducted by Statistics Canada that uses real data, such as actual payroll tax filings, and hence is regarded as more accurate.
As of March, the Survey of Employment, Payroll and Hours (SEPH) has shown that Canada’s labour market has been losing 5,400 jobs a month, a period that the household survey has shown jobs growth averaging 19,000 a month. Taken cumulatively, that means over the six month period SEPH has been tracking a slow drain of jobs, while the household survey has been tracking a healthy gain.
The two surveys are not identical because the SEPH does not include jobs in the agriculture sector, nor does it count self-employment, but the two should generally agree over time.
Gregory says economists have come to expect “payback” for any outsized jobs growth or contraction in the household survey, so much so that they calculate a correction in their forecasts for the next month.
But it’s the headline number that everyone fixates on, regardless of whether it can be fairly considered exact, especially when it differs substantially to economic forecasts.
“We get an earful from clients internally and externally. How can you guys be so wrong?’ ” Gregory shouts, mimicking the complaints. “Well, the range (survey error factor) is large,” he explains.
And unlike economists in the U.S., Canadian forecasters have no data to point them in the right direction leading up to the report, such a weakly initial claims and manufacturing surveys that U.S. economists count on.
To make the situation even stranger — and more confusing — when Canada reports it’s unemployment rate at the beginning of each month, it does so using the household survey results. When the U.S. reports, usually at the same time, it is using its variation of the payroll survey.