Skip to main content

Why forecasting Canada's jobs report is so difficult

Every month, notable economists estimate the change in employment. Take your best shot at beating them. (Spoiler: You'll beat them – maybe even on the first try)

Canada's jobs report is among the most anticipated releases on the economic calendar, holding the power to move markets and inspire tough questions for elected officials.

It's also incredibly difficult to forecast.

Every month, several economists estimate the nation's jobless rate and the net change in employment, key figures in Statistics Canada's Labour Force Survey. And at first glance, they appear to have a lousy record of forecasting the second measure.

Story continues below advertisement

Let's take a look at March of this year.

Twenty-one economists were surveyed by Bloomberg on that month's figures. The median estimate was a net gain of 10,000 positions for the month, with a low forecast of no change in employment and a high of 25,000 jobs added.

The result: a gain of nearly 41,000 positions.

For the economists, this is a fairly typical outcome.

Since 2000, the net change in employment has fallen outside the range of economist estimates 54 per cent of the time.

In fact, there's a good chance you can out-forecast the forecasters. Yes, you.

Let’s play a game.

Take your best shot at guessing the monthly change in employment. If you don’t succeed at first, keep trying: we guarantee you’ll eventually beat the economists. Below you’ll find 11 consecutive months of net changes in employment from a random period between 2000 and now. Make your prediction for the 12th month.

Instructions

Use your mouse or finger to drag the 12th bar up or down until you’re happy with your guess. We’ve provided the range of economist estimates to get you started.

Your current guess: 0 thousand jobs

Submit guess

Results

Your guess was . Statscan’s figure was .

Your estimate was closer than economists.

Play again

Congratulations: You probably just beat some economists at their own game.

Story continues below advertisement

The bad news: This doesn't mean you're a better prognosticator.

It turns out, the economists' forecasting record is not an indictment of their abilities, and to understand why, it's imperative to know what Statscan is reporting every month.

There's a simple explanation for why the monthly jobs-change figure is so difficult to predict:

"Because it's the result of a poll," says Avery Shenfeld, chief economist at CIBC Capital Markets.

What Statscan releases is not the actual change in employment.

Rather, the figure is an estimate, adjusted for seasonal factors, with a sizable sampling error. It can also be volatile, often swinging from positive to negative and vice versa. Here's what the result looks like over time:

Story continues below advertisement

Net monthly change in employment

In thousands of jobs

100

50

0

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

50

0

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

50

0

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

50

0

50

100

150

2000

2004

2008

2012

THE GLOBE AND MAIL SOURCE: STATSCAN

For these reasons, the monthly figure is really tough to pin down.

So if you beat the economists, you got lucky. (Or, you figured out that placing your guess within the economist range guaranteed you'd beat at least one forecaster.)

To produce the jobs report, Statscan conducts a survey of roughly 56,000 households – a sliver of the population. The results are subject to sampling error, or a standard error.

Let's look again at March. The economy churned out a net gain of 40,600 positions that month, and the standard error associated with the monthly change was 29,500.

With that in mind, here's another way of describing the month:

There's a 68-per-cent chance the true value of the monthly change in employment fell within one standard error of the estimate in either direction, or between net gains of 11,100 and 70,100. Also, there's a 95-per-cent chance the true value fell within two standard errors of the estimate, or between a net loss of 18,400 positions and a gain of 99,600.

That's right: it's entirely possible the labour market lost jobs in March.

The fact is, month-to-month fluctuations in the jobs report come with a big asterisk.

But that's not to suggest Labour Force Survey results are useless.

"More statistically precise inferences about the pace of employment growth can be obtained by combining information from several consecutive months," Statscan explains. "[T]he standard errors associated with the difference in the average monthly change in employment calculated over periods of three, six and 12 months are smaller than the standard error based solely on the current and previous reference months."

Taking a longer-term view of employment also smoothes out monthly volatility.

For instance, here's what the monthly jobs-change figure looks like versus a moving average:

Net monthly change in employment

In thousands of jobs

100

Net change

50

0

Seven-month

moving average

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

Net change

50

0

Seven-month

moving average

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

Net change

50

0

Seven-month

moving average

50

100

150

2000

2004

2008

2012

Net monthly change in employment

In thousands of jobs

100

Net change

50

0

Seven-month

moving average

50

100

150

2000

2004

2008

2012

THE GLOBE AND MAIL SOURCE: STATSCAN

The Labour Force Survey is not the sole measure of Canada's labour landscape, either.

Statscan's monthly Employment Insurance report looks at the number of Canadians receiving EI benefits, among other things. A big plus: it's not subject to sampling variability.

There's also the Survey of Employment, Payrolls and Hours, a monthly report whose results are widely considered more reliable than those of the Labour Force Survey.

But the jobs report gobbles up the attention, in large part because it's released just days after the month in question has ended, unlike other labour releases.

As such, there's a trade-off between timely data and more reliable data, says Mike Moffatt, an economics professor at Western University.

"And for whatever reason, as a society, we prefer the fast data rather than the good data," he says.

So why are economists taking a stab at the fast, noisy data?

"Because we're asked to," says Mr. Shenfeld.

"Investors want to have a view on what an economist thinks the next number might look like. We publish it, but don't place a lot of confidence in our ability to do this, and generally advise those we talk to not to put a lot of money behind any one of these forecasts."

Is there any hope of generating more accurate forecasts?

Douglas Porter, chief economist at BMO Nesbitt Burns, says "many have tried to come up with a better mousetrap in terms of estimating the Canadian employment numbers.

"But … getting that fine-tuned, as far as I'm aware, is impossible."

Note: The second chart features a seven-month moving average, using the three estimates immediately before and after the present result, as a means of visually eliminating a lag.

Source data for interactive are from Bloomberg

Report an error Editorial code of conduct
Due to technical reasons, we have temporarily removed commenting from our articles. We hope to have this fixed soon. Thank you for your patience. If you are looking to give feedback on our new site, please send it along to feedback@globeandmail.com. If you want to write a letter to the editor, please forward to letters@globeandmail.com.

Welcome to The Globe and Mail’s comment community. This is a space where subscribers can engage with each other and Globe staff. Non-subscribers can read and sort comments but will not be able to engage with them in any way. Click here to subscribe.

If you would like to write a letter to the editor, please forward it to letters@globeandmail.com. Readers can also interact with The Globe on Facebook and Twitter .

Welcome to The Globe and Mail’s comment community. This is a space where subscribers can engage with each other and Globe staff. Non-subscribers can read and sort comments but will not be able to engage with them in any way. Click here to subscribe.

If you would like to write a letter to the editor, please forward it to letters@globeandmail.com. Readers can also interact with The Globe on Facebook and Twitter .

Welcome to The Globe and Mail’s comment community. This is a space where subscribers can engage with each other and Globe staff.

We aim to create a safe and valuable space for discussion and debate. That means:

  • Treat others as you wish to be treated
  • Criticize ideas, not people
  • Stay on topic
  • Avoid the use of toxic and offensive language
  • Flag bad behaviour

Comments that violate our community guidelines will be removed.

Read our community guidelines here

Discussion loading ...

Latest Videos

To view this site properly, enable cookies in your browser. Read our privacy policy to learn more.
How to enable cookies