Meet the economists trying to publish economic data faster than StatsCan

Statistics Canada provides high-quality economic data, but it takes longer than its peers in other countries. Now a group of academics offer a faster alternative

Hare and tortoise with a rocket on its back

(Gandee Vasan/Getty)

Statistics Canada was in a reflective mood this week. “As 2017 marks the 150th anniversary of Confederation, we take a look back at the history of GDP as a means to examine the Canadian economy,” StatsCan said at the end of its latest report on Canada’s inflation-adjusted gross domestic product. The agency then presented a vignette from its archives.

The first release of the agency’s contemporary monthly GDP report was in the summer of 1981. This was a tough period in Canada’s economic history. The economy contracted 1.4% in July 1981 and another 0.5% in August, the two months covered in that initial report. The shock came from within. StatsCan said the decline was mainly attributable to a British Columbia forestry strike involving 60,000 workers.

Think about that number for a second. In December 2016, there were only 47,000 British Columbians employed in forestry, fishing, mining, and oil-and-gas combined. In the early 1980s, forestry and logging accounted for 1% of Canada’s GDP. Today that figure is about 0.2%. It explains why Prime Minister Justin Trudeau is so obsessed with the middle class. The economy his father oversaw has been transformed.

History is where StatsCan is strongest. However, it struggles to keep up with the demands of economists who prefer their data in real time. (Those people already know that GDP rose 0.4% in November from October, the fifth gain in six months. The November reading was 1.6% higher than a year earlier.) StatsCan reports on quarterly GDP eight weeks after the quarter ends, while the U.S. government waits only four weeks to release the first of three estimates. Bank of Canada Governor Stephen Poloz has urged StatsCan to provide monthly data on exports and imports of services, rather than including that information only in its quarterly reports on GDP. StatsCan takes almost two months to publish an assessment of the labour market based on a survey of companies. The U.S. Bureau of Labour Statistics processes a similar report in only a few weeks.

Randall Bartlett, chief economist at the University of Ottawa’s Institute of Fiscal Studies and Democracy (IFSD), calls StatsCan data the closest thing Canadians have to the “gospel truth.” Still, he says the Word needn’t take so long to come down from Tunney’s Pasture, the Ottawa neighbourhood where StatsCan is based. “Much of this data is released with a significant time lag relative to other developed countries,” Bartlett said.

StatsCan accepts that it could probably release data faster. But that would bring more revisions, said Jim Tebrake, director general of the macroeconomics branch. StatsCan proposed releasing quarterly GDP earlier, but the core users of its data, including the Finance Department, the Bank of Canada and provincial finance ministries, responded that they preferred to wait rather than deal with constant revisions. Other data series present collection issues. For example, data on trade in services some from a survey of companies. “It’s not realistic to bother businesses to give that information on a monthly basis,” Tebrake said. The agency is working on satisfying Poloz’s request, but “we need to be fairly innovative on how we do that,” Tebrake added.

The point of this exercise isn’t to beat up on Statistics Canada. It might be pokey, but its work is high class. Rather, the point is to flag IFSD’s attempt to offer a fix for data junkies who get the shakes waiting for StatsCan releases. The institute recently began publishing what amounts to the first “nowcast” of Canadian GDP that is readily available for public consumption.

If you haven’t heard of it, nowcasting is exactly what the name suggests. The proliferation of data and advances in economic modelling and computing power prompted some economists to wonder if they really needed to wait on statistics agencies to calculate headline figures such as GDP and employment. For about a decade, the profession has been growing increasingly confident that it doesn’t. Pretty much everyone who has an interest in the economic outlook is dabbling with nowcasting, crafting models that revise their forecasts the moment relevant data are released.

Yet few practitioners have been willing to make their models public. The Federal Reserve Bank of Atlanta has been producing a nowcast of U.S. economic growth called GDPNow since 2011. The New York Fed started publishing its own nowcast of GDP in April last year. But that’s about it as far as institutions go. Some banks release their nowcasts, but irregularly. The Bank of Canada has published a fair amount of research on the subject, and it has an in-house model that it follows, said Josianne Ménard, a spokeswoman. “Nowcasting models are useful for summarizing large amounts of high-frequency data,” Ménard said. “However, it is important to note that they are but one input into a wider set of analyses.” Nowcasting models sometimes fail to adjust for “special factors” or other events that are out of line with the historical readings on which the models are based, she said. The Bank of Canada doesn’t publish its nowcast.

Enter Bartlett and the IFSD. Earlier this year, they introduced @NowcastCanada, which anyone with a Twitter account can follow. Bartlett built a model that mirrors StatsCan’s methodology for tallying quarterly GDP and that also includes some U.S. indicators, such as oil imports. (In April 2016, a couple of researchers at the Federal Reserve Board published a paper on nowcasting Canadian GDP, concluding that a model with U.S. indicators was more accurate that one without.) When the “outstanding” November GDP number was released on January 31, the IFSD nowcast for fourth-quarter GDP was revised to 1.8%. That Bank of Canada’s current estimate, released with its latest quarterly economic report on January 18, is 1.6%. The central bank will revise that estimate when it publishes its spring Monetary Policy Report on April 12.

The IFSD isn’t trying to put forecasters out of business. StatsCan will retain the final word, and the Bank of Canada’s outlook will always guide markets because that is the one on which interest rates will be set. But nowcasts might offer a check on the work of Bay Street and Wall Street economists; Bartlett said research suggests nowcasting produces better predictions than calculating the consensus view of private-sector analysts. If that is true, imagine how useful a consensus of nowcasts could be? But for now, Canada has but one. “We would welcome others to publish their nowcasts,” said Bartlett. “The more information out there, the merrier.”

This story was updated with comment from Statistics Canada spokesperson Jim tebrake on February 3, 2017.