The May GDP number came out today, so it's time for my quarterly exercise in trying to come up with a preliminary estimate for quarterly GDP growth a month before Statistics Canada makes its first announcement. As regular readers will no doubt be weary of being reminded, the estimate is based on the GDP numbers for the first two months of the quarter along with the information in the LFS release for the third month. The most recent exercise is here.

The number I get for the annualised growth rate for GDP in the first quarter of 2012 is **1.6%**.

Here is the track record of these preliminary estimates:

Quarter WCI estimate First StatsCan

releaseLatest data 2009Q1 -6.9% -5.4% -7.9% 2009Q2 -3.4% -3.4% -3.7% 2009Q3 -0.4% 0.4% 1.7% 2009Q4 4.0% 5.0% 5.0% 2010Q1 5.5% 6.1% 5.6% 2010Q2 2.7% 2.0% 2.3% 2010Q3 1.5% 1.0% 2.5% 2010Q4 1.9% 3.3% 3.1% 2011Q1 3.8% 3.9% 3.6% 2011Q2 0.1% -0.4% -1.0% 2011Q3 3.1% 3.5% 4.5% 2011Q4 1.3% 1.8% 1.9% 2012Q1 1.6% 1.9% 1.9% 2012Q2 1.6%

My principal components based model is also flashing 1.6

Posted by: withheld | July 31, 2012 at 10:22 AM

Hello Steven:

Took a summary of the estimates. Over the period 2009Q1 to 2012Q1, the average WCI estimate is 1.14%, First Statscan is 1.52 and latest data is 1.50. In terms of median over same period, WCI estimate is 1.60%, First Statscan is 1.90% and Latest data is 2.30%.

Posted by: Livio Di Matteo | July 31, 2012 at 03:04 PM

I did a quick regression of the revisions on the WCI estimate; looks significant at the 10% level (and with only 13 obs!)

I'll leave it to Stephen to calculate the posterior odds.....

Posted by: Simon van Norden | July 31, 2012 at 06:01 PM

A few more regression results (with the Robusterrors option in RATS). Actually, Stephen might be doing better than StatCan does

one month later.1) If you regress StatCan's current estimate on the WCI (Stephen's) estimate and StatCan's preliminary estimate, they each get weights of around 0.5. (Ideally, if StatCan was doing the best possible, their estimate should get a weight of 1.0 and Stephen's should get a weight of zero.)

2) The difference between Stephen's estimate and StatCan's prelimnary number

helps predict StatCan's subsequent revision.(Chi-sq(1) stat of 12.5) That should imply that Stephen is adding info beyond what StatCan reports one month later.Caution: these results are based on 14 observations (and asymptotic theory for the hypothesis tests.)

Posted by: Simon van Norden | July 31, 2012 at 07:57 PM

Correction: the evidence for that 2nd claim is weaker (results above were from the wrong regression) - significant at about the 8% level.

Posted by: Simon van Norden | August 01, 2012 at 10:33 AM