StatsCan't

Not just a rounding error, then. With its face completely obscured by egg, Statistics Canada announced this morning that the economy added 42,000 jobs in July, and not the nugatory 200 it reported a couple of weeks ago. Seems a "human error" somewhere inside the agency led to a cohort of folks who were actually working being reported as unemployed.

It's not just StatsCan wearing egg-based makeup today, of course.  All of those commentators, including your humble scribe, who pounced on the original report as evidence of the Canadian economy's worrying underperformance don't come out of this very well, either. At the same time, it's hard to argue that even today's massively revised report offers a lot of comfort. Most of the jobs the economy is creating are part-time, and the overall performance of the Canadian economy in creating jobs is lagging well behind that of the US economy, whereas it normally tracks it closely. There's still every reason to believe that the weakness in Canada is structural rather than cyclical.

StatsCan is promising a full inquiry into the whole shambles in an effort to ensure it never happens again.  Here's a suggestion. In the past I've moaned about one of the more sterile duties of the business economist: "predicting" what the key data releases will be before they actually appear. (How can you call it "predicting" when the actual event has already happened?)  Even so, in the week or so before a key release like the employment data appears, a clear market consensus generally forms about what the number is likely to be. If StatsCan realizes that its number is vastly different from the consensus, which was certainly the case with the July jobs data, that should in and of itself trigger a special review of the data before they are released. You can never eliminate human error, but a procedure like this would reduce the risk of causing significant damage.