StatsCan's Recent Spate of Mistakes
Governments and businesses thrive on timely and accurate statistics. That's why a series of embarrassing recent snafus by STATISTICS CANADA, the agency charged with measuring the pulse of the nation, is so troubling. The biggest flub came in August, when StatsCan admitted it had underestimated five years of inflation data. But this year, there have been other incidents in which erroneous figures were released or computer systems failed to work properly. The fumbles come atop a long-standing complaint from economists and traders that StatsCan is infuriatingly slow at getting some of its numbers out. The lag time might have been grudgingly accepted when the quality of the product was impeccable. Now, the worry is, StatsCan is too late and too often wrong.
Take, for instance, retail trade, an estimate of how we spend our loonies. New sales data comes out each month. On Oct. 23, StatsCan released its latest figures, which dated back to August. That's downright poky compared to the number crunchers down south. Ten days earlier, the U.S. Department of Commerce had already cranked out its estimate of American retail spending for September, offering a far more up-to-date window into the minds of U.S. consumers.
Nobody expects StatsCan to count every Buick, beer and bread loaf sold in the country. Instead, it surveys thousands of businesses each month and extrapolates its results into national figures. Naturally, there are errors, and the U.S. Census Bureau certainly has its share of gaffes. When put side by side, though, it's not clear StatsCan is much more accurate than the Census Bureau for all the extra time it takes to crunch the Canadian numbers. "It's a trade-off between timeliness and accuracy," says Martin Barnes, editor of BCA Research, who focuses on the U.S. economy and relies heavily on Census Bureau data. "People put a high premium on timeliness."
StatsCan enjoys a solid reputation for having reliable numbers, compared to other developed nations like Germany and Japan. But the agency has taken its lumps over the last year. Aside from the inflation gaffe in August, which StatsCan blamed on a software error, last April the agency fessed up to undervaluing worker output in 2005 because they overlooked the fact 2004 was a leap year. Then, in September, a computer glitch blocked economists from accessing certain mercantile trade data. StatsCan's plodding production schedule makes such blunders more glaring. "The complaint out there from international market participants isn't so much that they're too sloppy," says one economist. "It's that they're too slow."
StatsCan has launched a review of how it certifies and disseminates monthly data, and the final report, due in January, will identify "weak points" in its system so they can be corrected, says Louis Marc Ducharme, director of the prices division at the agency. But on the issue of timeliness, he says the agency has no plans to change its tactics. For one thing, the agency uses larger samples for its surveys than the U.S., while publishing data in greater detail, all in an effort to be more accurate. "Historically, we in Canada have preferred to wait a few days more and have greater details and less revisions," says Ducharme. He points to data from the Organization for Economic Co-operation and Development, which tracks member nations' gross domestic product (GDP) and how often the figures are revised. Since 2000, Canada has had the smallest absolute revision of all G7 countries. "They show when you compare countries, we fit in the middle in terms of timeliness, but we're higher in terms of quality," he says. "It's a question of where you put your priority."
Maclean's November 13, 2006