It's time to update my series of posts (2009Q1, 2009Q2, 2009Q3, 2009Q4, 2010Q1) in which I try to take the GDP numbers from the first two months of a given quarter, mix them with the LFS numbers from the third month, and concoct a preliminary estimate for quarterly GDP growth. The BEA released theirs yesterday (2.4%); here's mine: 2.7%.
This is the sixth time I've done this, so let's all take it as read that that the result of this exercise is a probability distribution, not a single number. Its standard deviation is 0.5%, and the interquartile range is [2.3% , 3.0%].
Here is a table comparing the WCI 'backcasts' with the numbers produced by Statistics Canada:
Quarter WCI estimate First StatsCan
releaseLatest data 2009Q1 -6.9% -5.4% -7.3% 2009Q2 -3.4% -3.4% -2.8% 2009Q3 -0.4% 0.4% 0.9% 2009Q4 4.0% 5.0% 4.8% 2010Q1 5.5% 6.1% 6.1%
Given the track record of this methodology, it looks like we can say with a fair degree of confidence that the 2010Q2 numbers will be nowhere near as strong as the previous two quarters.
Worthwhile Canadian Initiative 
I would find it of interest to compare your chart with the forecasts that have been made by Mark Carney (or the Bank of Canada). Is that possible?
Stephen; Is that 0.5% std dev wrt StatCan preliminary estimates or current StatCan estimates?
Also, do preliminary estimates of the US growth rate add any explanitory power? Or changes in capacity utilisation survey measures?
It’s the predictive std deviation of the model: on the LHS is quarterly GDP, and on the RHS there is the first two months’ GDP numbers and the third month’s LFS data. The std deviation includes parameter uncertainty.
Next quarter I’ll see what happens when you add the preliminary US numbers.
Randy: I could add those to the table, but it wouldn’t be a fair comparison. I’m using data that forecasters didn’t have. The more appropriate comparison would be revisions to the BEA preliminary release – see the 2010Q1 post.
Average of Big 5 Banks and the BoC is 3.2% with a range of (2.7 (BMO) to 4.1 (TD)). My model is spitting out 3.2%.
Stephen; Thanks for the response, but I’m still not clear. Is it preliminary quarterly GDP on the LHS? or the current vintage estimates of quarterly GDP?
Oh – it’s current vintage. It would be nice to have vintage data for monthly GDP, but I suspect StatsCan isn’t about to start any new projects these days. A good project for a grad student, perhaps…